Authors: Jaeyeon Jang, Young-Ik Kim, Jisu Lim, Hyeonseong Lee
Abstract: Knowledge distillation (KD) is a technique used to transfer knowledge from an overparameterized teacher network to a less-parameterized student network, thereby minimizing the incurred performance loss. KD methods can be categorized into offline and online approaches. Offline KD leverages a powerful pretrained teacher network, while online KD allows the teacher network to be adjusted dynamically to enhance the learning effectiveness of the student network. Recently, it has been discovered that sharing the classifier of the teacher network can significantly boost the performance of the student network with only a minimal increase in the number of network parameters. Building on these insights, we propose adaptive teaching with a shared classifier (ATSC). In ATSC, the pretrained teacher network self-adjusts to better align with the learning needs of the student network based on its capabilities, and the student network benefits from the shared classifier, enhancing its performance. Additionally, we extend ATSC to environments with multiple teachers. We conduct extensive experiments, demonstrating the effectiveness of the proposed KD method. Our approach achieves state-of-the-art results on the CIFAR-100 and ImageNet datasets in both single-teacher and multiteacher scenarios, with only a modest increase in the number of required model parameters. The source code is publicly available at https://github.com/random2314235/ATSC.
Authors: Zhihang Yuan, Pu Lu, Hanling Zhang, Xuefei Ning, Linfeng Zhang, Tianchen Zhao, Shengen Yan, Guohao Dai, Yu Wang
Abstract: Diffusion Transformers (DiT) excel at image and video generation but face computational challenges due to self-attention's quadratic complexity. We propose DiTFastAttn, a novel post-training compression method to alleviate DiT's computational bottleneck. We identify three key redundancies in the attention computation during DiT inference: 1. spatial redundancy, where many attention heads focus on local information; 2. temporal redundancy, with high similarity between neighboring steps' attention outputs; 3. conditional redundancy, where conditional and unconditional inferences exhibit significant similarity. To tackle these redundancies, we propose three techniques: 1. Window Attention with Residual Caching to reduce spatial redundancy; 2. Temporal Similarity Reduction to exploit the similarity between steps; 3. Conditional Redundancy Elimination to skip redundant computations during conditional generation. To demonstrate the effectiveness of DiTFastAttn, we apply it to DiT, PixArt-Sigma for image generation tasks, and OpenSora for video generation tasks. Evaluation results show that for image generation, our method reduces up to 88\% of the FLOPs and achieves up to 1.6x speedup at high resolution generation.
Authors: Nhat Hoang-Xuan, Minh Vu, My T. Thai
Abstract: Providing textual concept-based explanations for neurons in deep neural networks (DNNs) is of importance in understanding how a DNN model works. Prior works have associated concepts with neurons based on examples of concepts or a pre-defined set of concepts, thus limiting possible explanations to what the user expects, especially in discovering new concepts. Furthermore, defining the set of concepts requires manual work from the user, either by directly specifying them or collecting examples. To overcome these, we propose to leverage multimodal large language models for automatic and open-ended concept discovery. We show that, without a restricted set of pre-defined concepts, our method gives rise to novel interpretable concepts that are more faithful to the model's behavior. To quantify this, we validate each concept by generating examples and counterexamples and evaluating the neuron's response on this new set of images. Collectively, our method can discover concepts and simultaneously validate them, providing a credible automated tool to explain deep neural networks.
Authors: George Cazenavette, Avneesh Sud, Thomas Leung, Ben Usman
Abstract: Due to the high potential for abuse of GenAI systems, the task of detecting synthetic images has recently become of great interest to the research community. Unfortunately, existing image-space detectors quickly become obsolete as new high-fidelity text-to-image models are developed at blinding speed. In this work, we propose a new synthetic image detector that uses features obtained by inverting an open-source pre-trained Stable Diffusion model. We show that these inversion features enable our detector to generalize well to unseen generators of high visual fidelity (e.g., DALL-E 3) even when the detector is trained only on lower fidelity fake images generated via Stable Diffusion. This detector achieves new state-of-the-art across multiple training and evaluation setups. Moreover, we introduce a new challenging evaluation protocol that uses reverse image search to mitigate stylistic and thematic biases in the detector evaluation. We show that the resulting evaluation scores align well with detectors' in-the-wild performance, and release these datasets as public benchmarks for future research.
Authors: Maria Pilligua, Nil Biescas, Javier Vazquez-Corral, Josep Llad\'os, Ernest Valveny, Sanket Biswas
Abstract: The rapid evolution of intelligent document processing systems demands robust solutions that adapt to diverse domains without extensive retraining. Traditional methods often falter with variable document types, leading to poor performance. To overcome these limitations, this paper introduces a text-graphic layer separation approach that enhances domain adaptability in document image restoration (DIR) systems. We propose LayeredDoc, which utilizes two layers of information: the first targets coarse-grained graphic components, while the second refines machine-printed textual content. This hierarchical DIR framework dynamically adjusts to the characteristics of the input document, facilitating effective domain adaptation. We evaluated our approach both qualitatively and quantitatively using a new real-world dataset, LayeredDocDB, developed for this study. Initially trained on a synthetically generated dataset, our model demonstrates strong generalization capabilities for the DIR task, offering a promising solution for handling variability in real-world data. Our code is accessible on GitHub.
Authors: Ahmed Abul Hasanaath, Hamzah Luqman, Raed Katib, Saeed Anwar
Abstract: Advances in deepfake research have led to the creation of almost perfect manipulations undetectable by human eyes and some deepfakes detection tools. Recently, several techniques have been proposed to differentiate deepfakes from realistic images and videos. This paper introduces a Frequency Enhanced Self-Blended Images (FSBI) approach for deepfakes detection. This proposed approach utilizes Discrete Wavelet Transforms (DWT) to extract discriminative features from the self-blended images (SBI) to be used for training a convolutional network architecture model. The SBIs blend the image with itself by introducing several forgery artifacts in a copy of the image before blending it. This prevents the classifier from overfitting specific artifacts by learning more generic representations. These blended images are then fed into the frequency features extractor to detect artifacts that can not be detected easily in the time domain. The proposed approach has been evaluated on FF++ and Celeb-DF datasets and the obtained results outperformed the state-of-the-art techniques with the cross-dataset evaluation protocol.
Authors: Weixi Feng, Jiachen Li, Michael Saxon, Tsu-jui Fu, Wenhu Chen, William Yang Wang
Abstract: Video generation has many unique challenges beyond those of image generation. The temporal dimension introduces extensive possible variations across frames, over which consistency and continuity may be violated. In this study, we move beyond evaluating simple actions and argue that generated videos should incorporate the emergence of new concepts and their relation transitions like in real-world videos as time progresses. To assess the Temporal Compositionality of video generation models, we propose TC-Bench, a benchmark of meticulously crafted text prompts, corresponding ground truth videos, and robust evaluation metrics. The prompts articulate the initial and final states of scenes, effectively reducing ambiguities for frame development and simplifying the assessment of transition completion. In addition, by collecting aligned real-world videos corresponding to the prompts, we expand TC-Bench's applicability from text-conditional models to image-conditional ones that can perform generative frame interpolation. We also develop new metrics to measure the completeness of component transitions in generated videos, which demonstrate significantly higher correlations with human judgments than existing metrics. Our comprehensive experimental results reveal that most video generators achieve less than 20% of the compositional changes, highlighting enormous space for future improvement. Our analysis indicates that current video generation models struggle to interpret descriptions of compositional changes and synthesize various components across different time steps.
Authors: Bing Li, Cheng Zheng, Wenxuan Zhu, Jinjie Mai, Biao Zhang, Peter Wonka, Bernard Ghanem
Abstract: While diffusion models have shown impressive performance in 2D image/video generation, diffusion-based Text-to-Multi-view-Video (T2MVid) generation remains underexplored. The new challenges posed by T2MVid generation lie in the lack of massive captioned multi-view videos and the complexity of modeling such multi-dimensional distribution. To this end, we propose a novel diffusion-based pipeline that generates high-quality multi-view videos centered around a dynamic 3D object from text. Specifically, we factor the T2MVid problem into viewpoint-space and time components. Such factorization allows us to combine and reuse layers of advanced pre-trained multi-view image and 2D video diffusion models to ensure multi-view consistency as well as temporal coherence for the generated multi-view videos, largely reducing the training cost. We further introduce alignment modules to align the latent spaces of layers from the pre-trained multi-view and the 2D video diffusion models, addressing the reused layers' incompatibility that arises from the domain gap between 2D and multi-view data. In support of this and future research, we further contribute a captioned multi-view video dataset. Experimental results demonstrate that our method generates high-quality multi-view videos, exhibiting vivid motions, temporal coherence, and multi-view consistency, given a variety of text prompts.
Authors: Ben Agro, Quinlan Sykora, Sergio Casas, Thomas Gilles, Raquel Urtasun
Abstract: Perceiving the world and forecasting its future state is a critical task for self-driving. Supervised approaches leverage annotated object labels to learn a model of the world -- traditionally with object detections and trajectory predictions, or temporal bird's-eye-view (BEV) occupancy fields. However, these annotations are expensive and typically limited to a set of predefined categories that do not cover everything we might encounter on the road. Instead, we learn to perceive and forecast a continuous 4D (spatio-temporal) occupancy field with self-supervision from LiDAR data. This unsupervised world model can be easily and effectively transferred to downstream tasks. We tackle point cloud forecasting by adding a lightweight learned renderer and achieve state-of-the-art performance in Argoverse 2, nuScenes, and KITTI. To further showcase its transferability, we fine-tune our model for BEV semantic occupancy forecasting and show that it outperforms the fully supervised state-of-the-art, especially when labeled data is scarce. Finally, when compared to prior state-of-the-art on spatio-temporal geometric occupancy prediction, our 4D world model achieves a much higher recall of objects from classes relevant to self-driving.
Authors: Abdel Rahman Alsabbagh, Omar Al-Kadi
Abstract: Generative Adversarial Networks (GANs) have exhibited noteworthy advancements across various applications, including medical imaging. While numerous state-of-the-art Deep Convolutional Neural Network (DCNN) architectures are renowned for their proficient feature extraction, this paper investigates their efficacy in the context of medical image deepfake detection. The primary objective is to effectively distinguish real from tampered or manipulated medical images by employing a comprehensive evaluation of 13 state-of-the-art DCNNs. Performance is assessed across diverse evaluation metrics, encompassing considerations of time efficiency and computational resource requirements. Our findings reveal that ResNet50V2 excels in precision and specificity, whereas DenseNet169 is distinguished by its accuracy, recall, and F1-score. We investigate the specific scenarios in which one model would be more favorable than another. Additionally, MobileNetV3Large offers competitive performance, emerging as the swiftest among the considered DCNN models while maintaining a relatively small parameter count. We also assess the latent space separability quality across the examined DCNNs, showing superiority in both the DenseNet and EfficientNet model families and entailing a higher understanding of medical image deepfakes. The experimental analysis in this research contributes valuable insights to the field of deepfake image detection in the medical imaging domain.
Authors: Fengyi Zhang, Tianjun Zhang, Lin Zhang, Helen Huang, Yadan Luo
Abstract: The field of novel-view synthesis has recently witnessed the emergence of 3D Gaussian Splatting, which represents scenes in a point-based manner and renders through rasterization. This methodology, in contrast to Radiance Fields that rely on ray tracing, demonstrates superior rendering quality and speed. However, the explicit and unstructured nature of 3D Gaussians poses a significant storage challenge, impeding its broader application. To address this challenge, we introduce the Gaussian-Forest modeling framework, which hierarchically represents a scene as a forest of hybrid 3D Gaussians. Each hybrid Gaussian retains its unique explicit attributes while sharing implicit ones with its sibling Gaussians, thus optimizing parameterization with significantly fewer variables. Moreover, adaptive growth and pruning strategies are designed, ensuring detailed representation in complex regions and a notable reduction in the number of required Gaussians. Extensive experiments demonstrate that Gaussian-Forest not only maintains comparable speed and quality but also achieves a compression rate surpassing 10 times, marking a significant advancement in efficient scene modeling. Codes are available at https://github.com/Xian-Bei/GaussianForest.
Authors: Xuannan Liu, Zekun Li, Peipei Li, Shuhan Xia, Xing Cui, Linzhi Huang, Huaibo Huang, Weihong Deng, Zhaofeng He
Abstract: Current multimodal misinformation detection (MMD) methods often assume a single source and type of forgery for each sample, which is insufficient for real-world scenarios where multiple forgery sources coexist. The lack of a benchmark for mixed-source misinformation has hindered progress in this field. To address this, we introduce MMFakeBench, the first comprehensive benchmark for mixed-source MMD. MMFakeBench includes 3 critical sources: textual veracity distortion, visual veracity distortion, and cross-modal consistency distortion, along with 12 sub-categories of misinformation forgery types. We further conduct an extensive evaluation of 6 prevalent detection methods and 15 large vision-language models (LVLMs) on MMFakeBench under a zero-shot setting. The results indicate that current methods struggle under this challenging and realistic mixed-source MMD setting. Additionally, we propose an innovative unified framework, which integrates rationales, actions, and tool-use capabilities of LVLM agents, significantly enhancing accuracy and generalization. We believe this study will catalyze future research into more realistic mixed-source multimodal misinformation and provide a fair evaluation of misinformation detection methods.
Authors: Zhengrui Xu, Guan'an Wang, Xiaowen Huang, Jitao Sang
Abstract: In this paper, we propose a novel Denoising Model for Representation Learning and take Person Re-Identification (ReID) as a benchmark task, named DenoiseReID, to improve feature discriminative with joint feature extraction and denoising. In the deep learning epoch, backbones which consists of cascaded embedding layers (e.g. convolutions or transformers) to progressively extract useful features, becomes popular. We first view each embedding layer in a backbone as a denoising layer, processing the cascaded embedding layers as if we are recursively denoise features step-by-step. This unifies the frameworks of feature extraction and feature denoising, where the former progressively embeds features from low-level to high-level, and the latter recursively denoises features step-by-step. Then we design a novel Feature Extraction and Feature Denoising Fusion Algorithm (FEFDFA) and \textit{theoretically demonstrate} its equivalence before and after fusion. FEFDFA merges parameters of the denoising layers into existing embedding layers, thus making feature denoising computation-free. This is a label-free algorithm to incrementally improve feature also complementary to the label if available. Besides, it enjoys two advantages: 1) it's a computation-free and label-free plugin for incrementally improving ReID features. 2) it is complementary to the label if the label is available. Experimental results on various tasks (large-scale image classification, fine-grained image classification, image retrieval) and backbones (transformers and convolutions) show the scalability and stability of our method. Experimental results on 4 ReID datasets and various of backbones show the stability and impressive improvements. We also extend the proposed method to large-scale (ImageNet) and fine-grained (e.g. CUB200) classification tasks, similar improvements are proven.
Authors: Mohammed Abdul Hafeez Khan, Parth Ganeriwala, Siddhartha Bhattacharyya, Natasha Neogi, Raja Muthalagu
Abstract: Labels are the cornerstone of supervised machine learning algorithms. Most visual recognition methods are fully supervised, using bounding boxes or pixel-wise segmentations for object localization. Traditional labeling methods, such as crowd-sourcing, are prohibitive due to cost, data privacy, amount of time, and potential errors on large datasets. To address these issues, we propose a novel annotation framework, Advanced Line Identification and Notation Algorithm (ALINA), which can be used for labeling taxiway datasets that consist of different camera perspectives and variable weather attributes (sunny and cloudy). Additionally, the CIRCular threshoLd pixEl Discovery And Traversal (CIRCLEDAT) algorithm has been proposed, which is an integral step in determining the pixels corresponding to taxiway line markings. Once the pixels are identified, ALINA generates corresponding pixel coordinate annotations on the frame. Using this approach, 60,249 frames from the taxiway dataset, AssistTaxi have been labeled. To evaluate the performance, a context-based edge map (CBEM) set was generated manually based on edge features and connectivity. The detection rate after testing the annotated labels with the CBEM set was recorded as 98.45%, attesting its dependability and effectiveness.
Authors: Wenjie Wang, Yehao Lu, Guangcong Zheng, Shuigen Zhan, Xiaoqing Ye, Zichang Tan, Jingdong Wang, Gaoang Wang, Xi Li
Abstract: Vision-based roadside 3D object detection has attracted rising attention in autonomous driving domain, since it encompasses inherent advantages in reducing blind spots and expanding perception range. While previous work mainly focuses on accurately estimating depth or height for 2D-to-3D mapping, ignoring the position approximation error in the voxel pooling process. Inspired by this insight, we propose a novel voxel pooling strategy to reduce such error, dubbed BEVSpread. Specifically, instead of bringing the image features contained in a frustum point to a single BEV grid, BEVSpread considers each frustum point as a source and spreads the image features to the surrounding BEV grids with adaptive weights. To achieve superior propagation performance, a specific weight function is designed to dynamically control the decay speed of the weights according to distance and depth. Aided by customized CUDA parallel acceleration, BEVSpread achieves comparable inference time as the original voxel pooling. Extensive experiments on two large-scale roadside benchmarks demonstrate that, as a plug-in, BEVSpread can significantly improve the performance of existing frustum-based BEV methods by a large margin of (1.12, 5.26, 3.01) AP in vehicle, pedestrian and cyclist.
Authors: Shubhankar Borse, Shreya Kadambi, Nilesh Prasad Pandey, Kartikeya Bhardwaj, Viswanath Ganapathy, Sweta Priyadarshi, Risheek Garrepalli, Rafael Esteves, Munawar Hayat, Fatih Porikli
Abstract: While Low-Rank Adaptation (LoRA) has proven beneficial for efficiently fine-tuning large models, LoRA fine-tuned text-to-image diffusion models lack diversity in the generated images, as the model tends to copy data from the observed training samples. This effect becomes more pronounced at higher values of adapter strength and for adapters with higher ranks which are fine-tuned on smaller datasets. To address these challenges, we present FouRA, a novel low-rank method that learns projections in the Fourier domain along with learning a flexible input-dependent adapter rank selection strategy. Through extensive experiments and analysis, we show that FouRA successfully solves the problems related to data copying and distribution collapse while significantly improving the generated image quality. We demonstrate that FouRA enhances the generalization of fine-tuned models thanks to its adaptive rank selection. We further show that the learned projections in the frequency domain are decorrelated and prove effective when merging multiple adapters. While FouRA is motivated for vision tasks, we also demonstrate its merits for language tasks on the GLUE benchmark.
Authors: Mingwang Xu, Hui Li, Qingkun Su, Hanlin Shang, Liwei Zhang, Ce Liu, Jingdong Wang, Luc Van Gool, Yao Yao, Siyu Zhu
Abstract: The field of portrait image animation, driven by speech audio input, has experienced significant advancements in the generation of realistic and dynamic portraits. This research delves into the complexities of synchronizing facial movements and creating visually appealing, temporally consistent animations within the framework of diffusion-based methodologies. Moving away from traditional paradigms that rely on parametric models for intermediate facial representations, our innovative approach embraces the end-to-end diffusion paradigm and introduces a hierarchical audio-driven visual synthesis module to enhance the precision of alignment between audio inputs and visual outputs, encompassing lip, expression, and pose motion. Our proposed network architecture seamlessly integrates diffusion-based generative models, a UNet-based denoiser, temporal alignment techniques, and a reference network. The proposed hierarchical audio-driven visual synthesis offers adaptive control over expression and pose diversity, enabling more effective personalization tailored to different identities. Through a comprehensive evaluation that incorporates both qualitative and quantitative analyses, our approach demonstrates obvious enhancements in image and video quality, lip synchronization precision, and motion diversity. Further visualization and access to the source code can be found at: https://fudan-generative-vision.github.io/hallo.
Authors: Chaoqin Huang, Haoyan Guan, Aofan Jiang, Yanfeng Wang, Michael Spratling, Xinchao Wang, Ya Zhang
Abstract: Most existing anomaly detection methods require a dedicated model for each category. Such a paradigm, despite its promising results, is computationally expensive and inefficient, thereby failing to meet the requirements for real-world applications. Inspired by how humans detect anomalies, by comparing a query image to known normal ones, this paper proposes a novel few-shot anomaly detection (FSAD) framework. Using a training set of normal images from various categories, registration, aiming to align normal images of the same categories, is leveraged as the proxy task for self-supervised category-agnostic representation learning. At test time, an image and its corresponding support set, consisting of a few normal images from the same category, are supplied, and anomalies are identified by comparing the registered features of the test image to its corresponding support image features. Such a setup enables the model to generalize to novel test categories. It is, to our best knowledge, the first FSAD method that requires no model fine-tuning for novel categories: enabling a single model to be applied to all categories. Extensive experiments demonstrate the effectiveness of the proposed method. Particularly, it improves the current state-of-the-art for FSAD by 11.3% and 8.3% on the MVTec and MPDD benchmarks, respectively. The source code is available at https://github.com/Haoyan-Guan/CAReg.
Authors: Zhengqi Zhao, Xiaohu Huang, Hao Zhou, Kun Yao, Errui Ding, Jingdong Wang, Xinggang Wang, Wenyu Liu, Bin Feng
Abstract: The key to action counting is accurately locating each video's repetitive actions. Instead of estimating the probability of each frame belonging to an action directly, we propose a dual-branch network, i.e., SkimFocusNet, working in a two-step manner. The model draws inspiration from empirical observations indicating that humans typically engage in coarse skimming of entire sequences to grasp the general action pattern initially, followed by a finer, frame-by-frame focus to determine if it aligns with the target action. Specifically, SkimFocusNet incorporates a skim branch and a focus branch. The skim branch scans the global contextual information throughout the sequence to identify potential target action for guidance. Subsequently, the focus branch utilizes the guidance to diligently identify repetitive actions using a long-short adaptive guidance (LSAG) block. Additionally, we have observed that videos in existing datasets often feature only one type of repetitive action, which inadequately represents real-world scenarios. To more accurately describe real-life situations, we establish the Multi-RepCount dataset, which includes videos containing multiple repetitive motions. On Multi-RepCount, our SkimFoucsNet can perform specified action counting, that is, to enable counting a particular action type by referencing an exemplary video. This capability substantially exhibits the robustness of our method. Extensive experiments demonstrate that SkimFocusNet achieves state-of-the-art performances with significant improvements. We also conduct a thorough ablation study to evaluate the network components. The source code will be published upon acceptance.
Authors: Manish Kumar Singh, Rajeev Yasarla, Hong Cai, Mingu Lee, Fatih Porikli
Abstract: In this paper, we propose a novel token selective attention approach, ToSA, which can identify tokens that need to be attended as well as those that can skip a transformer layer. More specifically, a token selector parses the current attention maps and predicts the attention maps for the next layer, which are then used to select the important tokens that should participate in the attention operation. The remaining tokens simply bypass the next layer and are concatenated with the attended ones to re-form a complete set of tokens. In this way, we reduce the quadratic computation and memory costs as fewer tokens participate in self-attention while maintaining the features for all the image patches throughout the network, which allows it to be used for dense prediction tasks. Our experiments show that by applying ToSA, we can significantly reduce computation costs while maintaining accuracy on the ImageNet classification benchmark. Furthermore, we evaluate on the dense prediction task of monocular depth estimation on NYU Depth V2, and show that we can achieve similar depth prediction accuracy using a considerably lighter backbone with ToSA.
Authors: Richard Boadu Antwi, Samuel Takyi, Kimollo Michael, Alican Karaer, Eren Erman Ozguven, Ren Moses, Maxim A. Dulebenets, Thobias Sando
Abstract: Efficient and current roadway geometry data collection is critical to transportation agencies in road planning, maintenance, design, and rehabilitation. Data collection methods are divided into land-based and aerial-based. Land-based methods for extensive highway networks are tedious, costly, pose safety risks. Therefore, there is the need for efficient, safe, and economical data acquisition methodologies. The rise of computer vision and object detection technologies have made automated extraction of roadway geometry features feasible. This study detects roadway features on Florida's public roads from high-resolution aerial images using AI. The developed model achieved an average accuracy of 80.4 percent when compared with ground truth data. The extracted roadway geometry data can be integrated with crash and traffic data to provide valuable insights to policymakers and roadway users.
Authors: Jiacong Hu, Jingwen Ye, Zunlei Feng, Jiazhen Yang, Shunyu Liu, Xiaotian Yu, Lingxiang Jia, Mingli Song
Abstract: Convolutional Neural Networks (CNNs) are well-known for their vulnerability to adversarial attacks, posing significant security concerns. In response to these threats, various defense methods have emerged to bolster the model's robustness. However, most existing methods either focus on learning from adversarial perturbations, leading to overfitting to the adversarial examples, or aim to eliminate such perturbations during inference, inevitably increasing computational burdens. Conversely, clean training, which strengthens the model's robustness by relying solely on clean examples, can address the aforementioned issues. In this paper, we align with this methodological stream and enhance its generalizability to unknown adversarial examples. This enhancement is achieved by scrutinizing the behavior of latent features within the network. Recognizing that a correct prediction relies on the correctness of the latent feature's pattern, we introduce a novel and effective Feature Pattern Consistency Constraint (FPCC) method to reinforce the latent feature's capacity to maintain the correct feature pattern. Specifically, we propose Spatial-wise Feature Modification and Channel-wise Feature Selection to enhance latent features. Subsequently, we employ the Pattern Consistency Loss to constrain the similarity between the feature pattern of the latent features and the correct feature pattern. Our experiments demonstrate that the FPCC method empowers latent features to uphold correct feature patterns even in the face of adversarial examples, resulting in inherent adversarial robustness surpassing state-of-the-art models.
Authors: Wenhui Xiao, Rodrigo Santa Cruz, David Ahmedt-Aristizabal, Olivier Salvado, Clinton Fookes, Leo Lebrat
Abstract: Neural Rendering representations have significantly contributed to the field of 3D computer vision. Given their potential, considerable efforts have been invested to improve their performance. Nonetheless, the essential question of selecting training views is yet to be thoroughly investigated. This key aspect plays a vital role in achieving high-quality results and aligns with the well-known tenet of deep learning: "garbage in, garbage out". In this paper, we first illustrate the importance of view selection by demonstrating how a simple rotation of the test views within the most pervasive NeRF dataset can lead to consequential shifts in the performance rankings of state-of-the-art techniques. To address this challenge, we introduce a unified framework for view selection methods and devise a thorough benchmark to assess its impact. Significant improvements can be achieved without leveraging error or uncertainty estimation but focusing on uniform view coverage of the reconstructed object, resulting in a training-free approach. Using this technique, we show that high-quality renderings can be achieved faster by using fewer views. We conduct extensive experiments on both synthetic datasets and realistic data to demonstrate the effectiveness of our proposed method compared with random, conventional error-based, and uncertainty-guided view selection.
Authors: Maor Dikter, Tsachi Blau, Chaim Baskin
Abstract: Concept bottleneck models (CBMs) have emerged as critical tools in domains where interpretability is paramount. These models rely on predefined textual descriptions, referred to as concepts, to inform their decision-making process and offer more accurate reasoning. As a result, the selection of concepts used in the model is of utmost significance. This study proposes \underline{\textbf{C}}onceptual \underline{\textbf{L}}earning via \underline{\textbf{E}}mbedding \underline{\textbf{A}}pproximations for \underline{\textbf{R}}einforcing Interpretability and Transparency, abbreviated as CLEAR, a framework for constructing a CBM for image classification. Using score matching and Langevin sampling, we approximate the embedding of concepts within the latent space of a vision-language model (VLM) by learning the scores associated with the joint distribution of images and concepts. A concept selection process is then employed to optimize the similarity between the learned embeddings and the predefined ones. The derived bottleneck offers insights into the CBM's decision-making process, enabling more comprehensive interpretations. Our approach was evaluated through extensive experiments and achieved state-of-the-art performance on various benchmarks. The code for our experiments is available at https://github.com/clearProject/CLEAR/tree/main
Authors: Tianle Zhang, Langtian Ma, Yuchen Yan, Yuchen Zhang, Kai Wang, Yue Yang, Ziyao Guo, Wenqi Shao, Yang You, Yu Qiao, Ping Luo, Kaipeng Zhang
Abstract: Recent text-to-video (T2V) technology advancements, as demonstrated by models such as Gen2, Pika, and Sora, have significantly broadened its applicability and popularity. Despite these strides, evaluating these models poses substantial challenges. Primarily, due to the limitations inherent in automatic metrics, manual evaluation is often considered a superior method for assessing T2V generation. However, existing manual evaluation protocols face reproducibility, reliability, and practicality issues. To address these challenges, this paper introduces the Text-to-Video Human Evaluation (T2VHE) protocol, a comprehensive and standardized protocol for T2V models. The T2VHE protocol includes well-defined metrics, thorough annotator training, and an effective dynamic evaluation module. Experimental results demonstrate that this protocol not only ensures high-quality annotations but can also reduce evaluation costs by nearly 50%. We will open-source the entire setup of the T2VHE protocol, including the complete protocol workflow, the dynamic evaluation component details, and the annotation interface code. This will help communities establish more sophisticated human assessment protocols.
Authors: Jiangshan Wang, Yue Ma, Jiayi Guo, Yicheng Xiao, Gao Huang, Xiu Li
Abstract: Video editing is an emerging task, in which most current methods adopt the pre-trained text-to-image (T2I) diffusion model to edit the source video in a zero-shot manner. Despite extensive efforts, maintaining the temporal consistency of edited videos remains challenging due to the lack of temporal constraints in the regular T2I diffusion model. To address this issue, we propose COrrespondence-guided Video Editing (COVE), leveraging the inherent diffusion feature correspondence to achieve high-quality and consistent video editing. Specifically, we propose an efficient sliding-window-based strategy to calculate the similarity among tokens in the diffusion features of source videos, identifying the tokens with high correspondence across frames. During the inversion and denoising process, we sample the tokens in noisy latent based on the correspondence and then perform self-attention within them. To save GPU memory usage and accelerate the editing process, we further introduce the temporal-dimensional token merging strategy, which can effectively reduce redundancy. COVE can be seamlessly integrated into the pre-trained T2I diffusion model without the need for extra training or optimization. Extensive experiment results demonstrate that COVE achieves the start-of-the-art performance in various video editing scenarios, outperforming existing methods both quantitatively and qualitatively. The code will be release at https://github.com/wangjiangshan0725/COVE
Authors: Nabil Ibtehaz, Ning Yan, Masood Mortazavi, Daisuke Kihara
Abstract: Modern vision transformers leverage visually inspired local interaction between pixels through attention computed within window or grid regions, in contrast to the global attention employed in the original ViT. Regional attention restricts pixel interactions within specific regions, while sparse attention disperses them across sparse grids. These differing approaches pose a challenge between maintaining hierarchical relationships vs. capturing a global context. In this study, drawing inspiration from atrous convolution, we propose Atrous Attention, a blend of regional and sparse attention that dynamically integrates both local and global information while preserving hierarchical structures. Based on this, we introduce a versatile, hybrid vision transformer backbone called ACC-ViT, tailored for standard vision tasks. Our compact model achieves approximately 84% accuracy on ImageNet-1K with fewer than 28.5 million parameters, outperforming the state-of-the-art MaxViT by 0.42% while requiring 8.4% fewer parameters.
Authors: Jiahao Qin
Abstract: Feature alignment serves as the primary mechanism for fusing multimodal data. We put forth a feature alignment approach that achieves full integration of multimodal information. This is accomplished via an alternating process of shifting and expanding feature representations across modalities to obtain a consistent unified representation in a joint feature space. The proposed technique can reliably capture high-level interplay between features originating from distinct modalities. Consequently, substantial gains in multimodal learning performance are attained. Additionally, we demonstrate the superiority of our approach over other prevalent multimodal fusion schemes on a range of tasks. Extensive experimental evaluation conducted on multimodal datasets comprising time series, image, and text demonstrates that our method achieves state-of-the-art results.
Authors: Yuan-Ming Li, Wei-Jin Huang, An-Lan Wang, Ling-An Zeng, Jing-Ke Meng, Wei-Shi Zheng
Abstract: We present EgoExo-Fitness, a new full-body action understanding dataset, featuring fitness sequence videos recorded from synchronized egocentric and fixed exocentric (third-person) cameras. Compared with existing full-body action understanding datasets, EgoExo-Fitness not only contains videos from first-person perspectives, but also provides rich annotations. Specifically, two-level temporal boundaries are provided to localize single action videos along with sub-steps of each action. More importantly, EgoExo-Fitness introduces innovative annotations for interpretable action judgement--including technical keypoint verification, natural language comments on action execution, and action quality scores. Combining all of these, EgoExo-Fitness provides new resources to study egocentric and exocentric full-body action understanding across dimensions of "what", "when", and "how well". To facilitate research on egocentric and exocentric full-body action understanding, we construct benchmarks on a suite of tasks (i.e., action classification, action localization, cross-view sequence verification, cross-view skill determination, and a newly proposed task of guidance-based execution verification), together with detailed analysis. Code and data will be available at https://github.com/iSEE-Laboratory/EgoExo-Fitness/tree/main.
URLs: https://github.com/iSEE-Laboratory/EgoExo-Fitness/tree/main.
Authors: Paul Melki (IMS), Lionel Bombrun (IMS), Boubacar Diallo (IMS), J\'er\^ome Dias (IMS), Jean-Pierre da Costa (IMS)
Abstract: The deployment of safe and trustworthy machine learning systems, and particularly complex black box neural networks, in real-world applications requires reliable and certified guarantees on their performance. The conformal prediction framework offers such formal guarantees by transforming any point into a set predictor with valid, finite-set, guarantees on the coverage of the true at a chosen level of confidence. Central to this methodology is the notion of the nonconformity score function that assigns to each example a measure of ''strangeness'' in comparison with the previously seen observations. While the coverage guarantees are maintained regardless of the nonconformity measure, the point predictor and the dataset, previous research has shown that the performance of a conformal model, as measured by its efficiency (the average size of the predicted sets) and its informativeness (the proportion of prediction sets that are singletons), is influenced by the choice of the nonconformity score function. The current work introduces the Penalized Inverse Probability (PIP) nonconformity score, and its regularized version RePIP, that allow the joint optimization of both efficiency and informativeness. Through toy examples and empirical results on the task of crop and weed image classification in agricultural robotics, the current work shows how PIP-based conformal classifiers exhibit precisely the desired behavior in comparison with other nonconformity measures and strike a good balance between informativeness and efficiency.
Authors: Zheng Dang, Jialu Huang, Fei Wang, Mathieu Salzmann
Abstract: Recent advances in deep learning such as neural radiance fields and implicit neural representations have significantly propelled the field of 3D reconstruction. However, accurately reconstructing objects with complex optical properties, such as metals and glass, remains a formidable challenge due to their unique specular and light-transmission characteristics. To facilitate the development of solutions to these challenges, we introduce the OpenMaterial dataset, comprising 1001 objects made of 295 distinct materials-including conductors, dielectrics, plastics, and their roughened variants- and captured under 723 diverse lighting conditions. To this end, we utilized physics-based rendering with laboratory-measured Indices of Refraction (IOR) and generated high-fidelity multiview images that closely replicate real-world objects. OpenMaterial provides comprehensive annotations, including 3D shape, material type, camera pose, depth, and object mask. It stands as the first large-scale dataset enabling quantitative evaluations of existing algorithms on objects with diverse and challenging materials, thereby paving the way for the development of 3D reconstruction algorithms capable of handling complex material properties.
Authors: Simon Bilik, Ilona Janakova, Adam Ligocki, Dominik Ficek, Karel Horak
Abstract: Many application from the bee colony health state monitoring could be efficiently solved using a computer vision techniques. One of such challenges is an efficient way for counting the number of incoming and outcoming bees, which could be used to further analyse many trends, such as the bee colony health state, blooming periods, or for investigating the effects of agricultural spraying. In this paper, we compare three methods for the automated bee counting over two own datasets. The best performing method is based on the ResNet-50 convolutional neural network classifier, which achieved accuracy of 87% over the BUT1 dataset and the accuracy of 93% over the BUT2 dataset.
Authors: Yue Xu, Kaizhi Yang, Jiebo Luo, Xuejin Chen
Abstract: 3D visual grounding is an emerging research area dedicated to making connections between the 3D physical world and natural language, which is crucial for achieving embodied intelligence. In this paper, we propose DASANet, a Dual Attribute-Spatial relation Alignment Network that separately models and aligns object attributes and spatial relation features between language and 3D vision modalities. We decompose both the language and 3D point cloud input into two separate parts and design a dual-branch attention module to separately model the decomposed inputs while preserving global context in attribute-spatial feature fusion by cross attentions. Our DASANet achieves the highest grounding accuracy 65.1% on the Nr3D dataset, 1.3% higher than the best competitor. Besides, the visualization of the two branches proves that our method is efficient and highly interpretable.
Authors: Chenyang Shi, Shasha Guo, Boyi Wei, Hanxiao Liu, Yibo Zhang, Ningfang Song, Jing Jin
Abstract: Event cameras are renowned for their high efficiency due to outputting a sparse, asynchronous stream of events. However, they are plagued by noisy events, especially in low light conditions. Denoising is an essential task for event cameras, but evaluating denoising performance is challenging. Label-dependent denoising metrics involve artificially adding noise to clean sequences, complicating evaluations. Moreover, the majority of these metrics are monotonic, which can inflate scores by removing substantial noise and valid events. To overcome these limitations, we propose the first label-free and non-monotonic evaluation metric, the area of the continuous contrast curve (AOCC), which utilizes the area enclosed by event frame contrast curves across different time intervals. This metric is inspired by how events capture the edge contours of scenes or objects with high temporal resolution. An effective denoising method removes noise without eliminating these edge-contour events, thus preserving the contrast of event frames. Consequently, contrast across various time ranges serves as a metric to assess denoising effectiveness. As the time interval lengthens, the curve will initially rise and then fall. The proposed metric is validated through both theoretical and experimental evidence.
Authors: Guodong Sun, Junjie Liu, Mingxuan Liu, Moyun Liu, Yang Zhang
Abstract: Self-supervised monocular depth estimation aims to infer depth information without relying on labeled data. However, the lack of labeled information poses a significant challenge to the model's representation, limiting its ability to capture the intricate details of the scene accurately. Prior information can potentially mitigate this issue, enhancing the model's understanding of scene structure and texture. Nevertheless, solely relying on a single type of prior information often falls short when dealing with complex scenes, necessitating improvements in generalization performance. To address these challenges, we introduce a novel self-supervised monocular depth estimation model that leverages multiple priors to bolster representation capabilities across spatial, context, and semantic dimensions. Specifically, we employ a hybrid transformer and a lightweight pose network to obtain long-range spatial priors in the spatial dimension. Then, the context prior attention is designed to improve generalization, particularly in complex structures or untextured areas. In addition, semantic priors are introduced by leveraging semantic boundary loss, and semantic prior attention is supplemented, further refining the semantic features extracted by the decoder. Experiments on three diverse datasets demonstrate the effectiveness of the proposed model. It integrates multiple priors to comprehensively enhance the representation ability, improving the accuracy and reliability of depth estimation. Codes are available at: \url{https://github.com/MVME-HBUT/MPRLNet}
Authors: Tuan Pham, Stephan Mandt
Abstract: Neural Radiance Fields (NeRFs) have emerged as powerful tools for capturing detailed 3D scenes through continuous volumetric representations. Recent NeRFs utilize feature grids to improve rendering quality and speed; however, these representations introduce significant storage overhead. This paper presents a novel method for efficiently compressing a grid-based NeRF model, addressing the storage overhead concern. Our approach is based on the non-linear transform coding paradigm, employing neural compression for compressing the model's feature grids. Due to the lack of training data involving many i.i.d scenes, we design an encoder-free, end-to-end optimized approach for individual scenes, using lightweight decoders. To leverage the spatial inhomogeneity of the latent feature grids, we introduce an importance-weighted rate-distortion objective and a sparse entropy model employing a masking mechanism. Our experimental results validate that our proposed method surpasses existing works in terms of grid-based NeRF compression efficacy and reconstruction quality.
Authors: Duong H. Le, Tuan Pham, Aniruddha Kembhavi, Stephan Mandt, Wei-Chiu Ma, Jiasen Lu
Abstract: We present Piva (Preserving Identity with Variational Score Distillation), a novel optimization-based method for editing images and 3D models based on diffusion models. Specifically, our approach is inspired by the recently proposed method for 2D image editing - Delta Denoising Score (DDS). We pinpoint the limitations in DDS for 2D and 3D editing, which causes detail loss and over-saturation. To address this, we propose an additional score distillation term that enforces identity preservation. This results in a more stable editing process, gradually optimizing NeRF models to match target prompts while retaining crucial input characteristics. We demonstrate the effectiveness of our approach in zero-shot image and neural field editing. Our method successfully alters visual attributes, adds both subtle and substantial structural elements, translates shapes, and achieves competitive results on standard 2D and 3D editing benchmarks. Additionally, our method imposes no constraints like masking or pre-training, making it compatible with a wide range of pre-trained diffusion models. This allows for versatile editing without needing neural field-to-mesh conversion, offering a more user-friendly experience.
Authors: Jamie Watson, Filippo Aleotti, Mohamed Sayed, Zawar Qureshi, Oisin Mac Aodha, Gabriel Brostow, Michael Firman, Sara Vicente
Abstract: Extracting planes from a 3D scene is useful for downstream tasks in robotics and augmented reality. In this paper we tackle the problem of estimating the planar surfaces in a scene from posed images. Our first finding is that a surprisingly competitive baseline results from combining popular clustering algorithms with recent improvements in 3D geometry estimation. However, such purely geometric methods are understandably oblivious to plane semantics, which are crucial to discerning distinct planes. To overcome this limitation, we propose a method that predicts multi-view consistent plane embeddings that complement geometry when clustering points into planes. We show through extensive evaluation on the ScanNetV2 dataset that our new method outperforms existing approaches and our strong geometric baseline for the task of plane estimation.
Authors: Fengyuan Zhang, Zhaopei Huang, Xinjie Zhang, Qin Jin
Abstract: Micro-expressions serve as essential cues for understanding individuals' genuine emotional states. Recognizing micro-expressions attracts increasing research attention due to its various applications in fields such as business negotiation and psychotherapy. However, the intricate and transient nature of micro-expressions poses a significant challenge to their accurate recognition. Most existing works either neglect temporal dependencies or suffer from redundancy issues in clip-level recognition. In this work, we propose a novel framework for micro-expression recognition, named the Adaptive Temporal Motion Guided Graph Convolution Network (ATM-GCN). Our framework excels at capturing temporal dependencies between frames across the entire clip, thereby enhancing micro-expression recognition at the clip level. Specifically, the integration of Adaptive Temporal Motion layers empowers our method to aggregate global and local motion features inherent in micro-expressions. Experimental results demonstrate that ATM-GCN not only surpasses existing state-of-the-art methods, particularly on the Composite dataset, but also achieves superior performance on the latest micro-expression dataset CAS(ME)$^3$.
Authors: Lincan Cai, Shuang Li, Wenxuan Ma, Jingxuan Kang, Binhui Xie, Zixun Sun, Chengwei Zhu
Abstract: Large-scale pretrained models have proven immensely valuable in handling data-intensive modalities like text and image. However, fine-tuning these models for certain specialized modalities, such as protein sequence and cosmic ray, poses challenges due to the significant modality discrepancy and scarcity of labeled data. In this paper, we propose an end-to-end method, PaRe, to enhance cross-modal fine-tuning, aiming to transfer a large-scale pretrained model to various target modalities. PaRe employs a gating mechanism to select key patches from both source and target data. Through a modality-agnostic Patch Replacement scheme, these patches are preserved and combined to construct data-rich intermediate modalities ranging from easy to hard. By gradually intermediate modality generation, we can not only effectively bridge the modality gap to enhance stability and transferability of cross-modal fine-tuning, but also address the challenge of limited data in the target modality by leveraging enriched intermediate modality data. Compared with hand-designed, general-purpose, task-specific, and state-of-the-art cross-modal fine-tuning approaches, PaRe demonstrates superior performance across three challenging benchmarks, encompassing more than ten modalities.
Authors: Yingying Wang
Abstract: The traditional ingle-scale U-Net often leads to the loss of spatial information during deblurring, which affects the deblurring accracy. Additionally, due to the convolutional method's limitation in capturing long-range dependencies, the quality of the recovered image is degraded. To address the above problems, an asymmetric multiple scales U-net based on self-attention (AMSA-UNet) is proposed to improve the accuracy and computational complexity. By introducing a multiple-scales U shape architecture, the network can focus on blurry regions at the global level and better recover image details at the local level. In order to overcome the limitations of traditional convolutional methods in capturing the long-range dependencies of information, a self-attention mechanism is introduced into the decoder part of the backbone network, which significantly increases the model's receptive field, enabling it to pay more attention to semantic information of the image, thereby producing more accurate and visually pleasing deblurred images. What's more, a frequency domain-based computation method was introduced to reduces the computation amount. The experimental results demonstrate that the proposed method exhibits significant improvements in both accuracy and speed compared to eight excellent methods
Authors: Gaochang Wu, Yapeng Zhang, Lan Deng, Jingxin Zhang, Tianyou Chai
Abstract: Fused Magnesium Furnace (FMF) is a crucial industrial equipment in the production of magnesia, and anomaly detection plays a pivotal role in ensuring its efficient, stable, and secure operation. Existing anomaly detection methods primarily focus on analyzing dominant anomalies using the process variables (such as arc current) or constructing neural networks based on abnormal visual features, while overlooking the intrinsic correlation of cross-modal information. This paper proposes a cross-modal Transformer (dubbed FmFormer), designed to facilitate anomaly detection in fused magnesium smelting processes by exploring the correlation between visual features (video) and process variables (current). Our approach introduces a novel tokenization paradigm to effectively bridge the substantial dimensionality gap between the 3D video modality and the 1D current modality in a multiscale manner, enabling a hierarchical reconstruction of pixel-level anomaly detection. Subsequently, the FmFormer leverages self-attention to learn internal features within each modality and bidirectional cross-attention to capture correlations across modalities. To validate the effectiveness of the proposed method, we also present a pioneering cross-modal benchmark of the fused magnesium smelting process, featuring synchronously acquired video and current data for over 2.2 million samples. Leveraging cross-modal learning, the proposed FmFormer achieves state-of-the-art performance in detecting anomalies, particularly under extreme interferences such as current fluctuations and visual occlusion caused by heavy water mist. The presented methodology and benchmark may be applicable to other industrial applications with some amendments. The benchmark will be released at https://github.com/GaochangWu/FMF-Benchmark.
Authors: Shivansh Chandra Tripathi, Rahul Garg
Abstract: The Facial Action Coding System (FACS) for studying facial expressions is manual and requires significant effort and expertise. This paper explores the use of automated techniques to generate Action Units (AUs) for studying facial expressions. We propose an unsupervised approach based on Principal Component Analysis (PCA) and facial keypoint tracking to generate data-driven AUs called PCA AUs using the publicly available DISFA dataset. The PCA AUs comply with the direction of facial muscle movements and are capable of explaining over 92.83 percent of the variance in other public test datasets (BP4D-Spontaneous and CK+), indicating their capability to generalize facial expressions. The PCA AUs are also comparable to a keypoint-based equivalence of FACS AUs in terms of variance explained on the test datasets. In conclusion, our research demonstrates the potential of automated techniques to be an alternative to manual FACS labeling which could lead to efficient real-time analysis of facial expressions in psychology and related fields. To promote further research, we have made code repository publicly available.
Authors: Pei Yang, Hai Ci, Yiren Song, Mike Zheng Shou
Abstract: Digital watermarking techniques are crucial for copyright protection and source identification of images, especially in the era of generative AI models. However, many existing watermarking methods, particularly content-agnostic approaches that embed fixed patterns regardless of image content, are vulnerable to steganalysis attacks that can extract and remove the watermark with minimal perceptual distortion. In this work, we categorize watermarking algorithms into content-adaptive and content-agnostic ones, and demonstrate how averaging a collection of watermarked images could reveal the underlying watermark pattern. We then leverage this extracted pattern for effective watermark removal under both graybox and blackbox settings, even when the collection contains multiple watermark patterns. For some algorithms like Tree-Ring watermarks, the extracted pattern can also forge convincing watermarks on clean images. Our quantitative and qualitative evaluations across twelve watermarking methods highlight the threat posed by steganalysis to content-agnostic watermarks and the importance of designing watermarking techniques resilient to such analytical attacks. We propose security guidelines calling for using content-adaptive watermarking strategies and performing security evaluation against steganalysis. We also suggest multi-key assignments as potential mitigations against steganalysis vulnerabilities.
Authors: Hamza Bouzid, Lahoucine Ballihi
Abstract: Facial expressions, vital in non-verbal human communication, have found applications in various computer vision fields like virtual reality, gaming, and emotional AI assistants. Despite advancements, many facial expression generation models encounter challenges such as low resolution (e.g., 32x32 or 64x64 pixels), poor quality, and the absence of background details. In this paper, we introduce FacEnhance, a novel diffusion-based approach addressing constraints in existing low-resolution facial expression generation models. FacEnhance enhances low-resolution facial expression videos (64x64 pixels) to higher resolutions (192x192 pixels), incorporating background details and improving overall quality. Leveraging conditional denoising within a diffusion framework, guided by a background-free low-resolution video and a single neutral expression high-resolution image, FacEnhance generates a video incorporating the facial expression from the low-resolution video performed by the individual with background from the neutral image. By complementing lightweight low-resolution models, FacEnhance strikes a balance between computational efficiency and desirable image resolution and quality. Extensive experiments on the MUG facial expression database demonstrate the efficacy of FacEnhance in enhancing low-resolution model outputs to state-of-the-art quality while preserving content and identity consistency. FacEnhance represents significant progress towards resource-efficient, high-fidelity facial expression generation, Renewing outdated low-resolution methods to up-to-date standards.
Authors: Tarun Khajuria, Braian Olmiro Dias, Jaan Aru
Abstract: Forming and using symbol-like structured representations for reasoning has been considered essential for generalising over novel inputs. The primary tool that allows generalisation outside training data distribution is the ability to abstract away irrelevant information into a compact form relevant to the task. An extreme form of such abstract representations is symbols. Humans make use of symbols to bind information while abstracting away irrelevant parts to utilise the information consistently and meaningfully. This work estimates the state of such structured representations in vision encoders. Specifically, we evaluate image encoders in large vision-language pre-trained models to address the question of which desirable properties their representations lack by applying the criteria of symbolic structured reasoning described for LLMs to the image models. We test the representation space of image encoders like VIT, BLIP, CLIP, and FLAVA to characterise the distribution of the object representations in these models. In particular, we create decoding tasks using multi-object scenes from the COCO dataset, relating the token space to its input content for various objects in the scene. We use these tasks to characterise the network's token and layer-wise information modelling. Our analysis highlights that the CLS token, used for the downstream task, only focuses on a few objects necessary for the trained downstream task. Still, other individual objects are well-modelled separately by the tokens in the network originating from those objects. We further observed a widespread distribution of scene information. This demonstrates that information is far more entangled in tokens than optimal for representing objects similar to symbols. Given these symbolic properties, we show the network dynamics that cause failure modes of these models on basic downstream tasks in a multi-object scene.
Authors: Basim Azam, Naveed Akhtar
Abstract: Kolmogorov-Arnold Networks (KANs) introduce a paradigm of neural modeling that implements learnable functions on the edges of the networks, diverging from the traditional node-centric activations in neural networks. This work assesses the applicability and efficacy of KANs in visual modeling, focusing on the image recognition task. We mainly analyze the performance and efficiency of different network architectures built using KAN concepts along with conventional building blocks of convolutional and linear layers, enabling a comparative analysis with the conventional models. Our findings are aimed at contributing to understanding the potential of KANs in computer vision, highlighting both their strengths and areas for further research. Our evaluation shows that whereas KAN-based architectures perform in-line with the original claims of KAN paper for performance and model-complexity in the case of simpler vision datasets like MNIST, the advantages seem to diminish even for slightly more complex datasets like CIFAR-10.
Authors: Chenwei Lin, Hanjia Lyu, Xian Xu, Jiebo Luo
Abstract: Large Vision-Language Models (LVLMs) have demonstrated outstanding performance in various general multimodal applications such as image recognition and visual reasoning, and have also shown promising potential in specialized domains. However, the application potential of LVLMs in the insurance domain-characterized by rich application scenarios and abundant multimodal data-has not been effectively explored. There is no systematic review of multimodal tasks in the insurance domain, nor a benchmark specifically designed to evaluate the capabilities of LVLMs in insurance. This gap hinders the development of LVLMs within the insurance domain. In this paper, we systematically review and distill multimodal tasks for four representative types of insurance: auto insurance, property insurance, health insurance, and agricultural insurance. We propose INS-MMBench, the first comprehensive LVLMs benchmark tailored for the insurance domain. INS-MMBench comprises a total of 2.2K thoroughly designed multiple-choice questions, covering 12 meta-tasks and 22 fundamental tasks. Furthermore, we evaluate multiple representative LVLMs, including closed-source models such as GPT-4o and open-source models like BLIP-2. This evaluation not only validates the effectiveness of our benchmark but also provides an in-depth performance analysis of current LVLMs on various multimodal tasks in the insurance domain. We hope that INS-MMBench will facilitate the further application of LVLMs in the insurance domain and inspire interdisciplinary development. Our dataset and evaluation code are available at https://github.com/FDU-INS/INS-MMBench.
Authors: Halil Bisgin, Andres Palechor, Mike Suter, Manuel G\"unther
Abstract: The goal for classification is to correctly assign labels to unseen samples. However, most methods misclassify samples with unseen labels and assign them to one of the known classes. Open-Set Classification (OSC) algorithms aim to maximize both closed and open-set recognition capabilities. Recent studies showed the utility of such algorithms on small-scale data sets, but limited experimentation makes it difficult to assess their performances in real-world problems. Here, we provide a comprehensive comparison of various OSC algorithms, including training-based (SoftMax, Garbage, EOS) and post-processing methods (Maximum SoftMax Scores, Maximum Logit Scores, OpenMax, EVM, PROSER), the latter are applied on features from the former. We perform our evaluation on three large-scale protocols that mimic real-world challenges, where we train on known and negative open-set samples, and test on known and unknown instances. Our results show that EOS helps to improve performance of almost all post-processing algorithms. Particularly, OpenMax and PROSER are able to exploit better-trained networks, demonstrating the utility of hybrid models. However, while most algorithms work well on negative test samples -- samples of open-set classes seen during training -- they tend to perform poorly when tested on samples of previously unseen unknown classes, especially in challenging conditions.
Authors: Injoon Hwang, Haewon Park, Youngwan Lee, Jooyoung Yang, SunJae Maeng
Abstract: Low-rank adaption (LoRA) is a prominent method that adds a small number of learnable parameters to the frozen pre-trained weights for parameter-efficient fine-tuning. Prompted by the question, ``Can we make its representation enough with LoRA weights solely at the final phase of finetuning without the pre-trained weights?'' In this work, we introduce Progressive Compression LoRA~(PC-LoRA), which utilizes low-rank adaptation (LoRA) to simultaneously perform model compression and fine-tuning. The PC-LoRA method gradually removes the pre-trained weights during the training process, eventually leaving only the low-rank adapters in the end. Thus, these low-rank adapters replace the whole pre-trained weights, achieving the goals of compression and fine-tuning at the same time. Empirical analysis across various models demonstrates that PC-LoRA achieves parameter and FLOPs compression rates of 94.36%/89.1% for vision models, e.g., ViT-B, and 93.42%/84.2% parameters and FLOPs compressions for language models, e.g., BERT.
Authors: Jiahao Nie, Gongjie Zhang, Wenbin An, Yap-Peng Tan, Alex C. Kot, Shijian Lu
Abstract: Despite the recent advancements in Multi-modal Large Language Models (MLLMs), understanding inter-object relations, i.e., interactions or associations between distinct objects, remains a major challenge for such models. This issue significantly hinders their advanced reasoning capabilities and is primarily due to the lack of large-scale, high-quality, and diverse multi-modal data essential for training and evaluating MLLMs. In this paper, we provide a taxonomy of inter-object relations and introduce Multi-Modal Relation Understanding (MMRel), a comprehensive dataset designed to bridge this gap by providing large-scale, high-quality and diverse data for studying inter-object relations with MLLMs. MMRel features three distinctive attributes: (i) It includes over 15K question-answer pairs, which are sourced from three distinct domains, ensuring large scale and high diversity; (ii) It contains a subset featuring highly unusual relations, on which MLLMs often fail due to hallucinations, thus are very challenging; (iii) It provides manually verified high-quality labels for inter-object relations. Thanks to these features, MMRel is ideal for evaluating MLLMs on relation understanding, as well as being used to fine-tune MLLMs to enhance relation understanding and even benefit overall performance in various vision-language tasks. Extensive experiments on various popular MLLMs validate the effectiveness of MMRel. Both MMRel dataset and the complete labeling scripts have been made publicly available.
Authors: Weijie Wei, Osman \"Ulger, Fatemeh Karimi Najadasl, Theo Gevers, Martin R. Oswald
Abstract: Existing perception methods for autonomous driving fall short of recognizing unknown entities not covered in the training data. Open-vocabulary methods offer promising capabilities in detecting any object but are limited by user-specified queries representing target classes. We propose AutoVoc3D, a framework for automatic object class recognition and open-ended segmentation. Evaluation on nuScenes showcases AutoVoc3D's ability to generate precise semantic classes and accurate point-wise segmentation. Moreover, we introduce Text-Point Semantic Similarity, a new metric to assess the semantic similarity between text and point cloud without eliminating novel classes.
Authors: Xintian Mao, Qingli Li, Yan Wang
Abstract: Despite the recent progress in enhancing the efficacy of image deblurring, the limited decoding capability constrains the upper limit of State-Of-The-Art (SOTA) methods. This paper proposes a pioneering work, Adaptive Patch Exiting Reversible Decoder (AdaRevD), to explore their insufficient decoding capability. By inheriting the weights of the well-trained encoder, we refactor a reversible decoder which scales up the single-decoder training to multi-decoder training while remaining GPU memory-friendly. Meanwhile, we show that our reversible structure gradually disentangles high-level degradation degree and low-level blur pattern (residual of the blur image and its sharp counterpart) from compact degradation representation. Besides, due to the spatially-variant motion blur kernels, different blur patches have various deblurring difficulties. We further introduce a classifier to learn the degradation degree of image patches, enabling them to exit at different sub-decoders for speedup. Experiments show that our AdaRevD pushes the limit of image deblurring, e.g., achieving 34.60 dB in PSNR on GoPro dataset.
Authors: Yucheng Han, Rui Wang, Chi Zhang, Juntao Hu, Pei Cheng, Bin Fu, Hanwang Zhang
Abstract: Recent advancements in image generation have enabled the creation of high-quality images from text conditions. However, when facing multi-modal conditions, such as text combined with reference appearances, existing methods struggle to balance multiple conditions effectively, typically showing a preference for one modality over others. To address this challenge, we introduce EMMA, a novel image generation model accepting multi-modal prompts built upon the state-of-the-art text-to-image (T2I) diffusion model, ELLA. EMMA seamlessly incorporates additional modalities alongside text to guide image generation through an innovative Multi-modal Feature Connector design, which effectively integrates textual and supplementary modal information using a special attention mechanism. By freezing all parameters in the original T2I diffusion model and only adjusting some additional layers, we reveal an interesting finding that the pre-trained T2I diffusion model can secretly accept multi-modal prompts. This interesting property facilitates easy adaptation to different existing frameworks, making EMMA a flexible and effective tool for producing personalized and context-aware images and even videos. Additionally, we introduce a strategy to assemble learned EMMA modules to produce images conditioned on multiple modalities simultaneously, eliminating the need for additional training with mixed multi-modal prompts. Extensive experiments demonstrate the effectiveness of EMMA in maintaining high fidelity and detail in generated images, showcasing its potential as a robust solution for advanced multi-modal conditional image generation tasks.
Authors: Wenlong Yu, Dongyue Chen, Qilong Wang, Qinghua Hu
Abstract: Fine-grained domain generalization (FGDG) is a more challenging task due to its small inter-class variations and relatively large intra-class disparities. When domain distribution changes, the fragility of subtle features leads to a pronounced deterioration in model performance.Nevertheless, humans inherently demonstrate the capacity for generalizing to out-of-distribution data, leveraging structured multi-granularity knowledge that emerges from discerning both the commonality and specificity within categories.Likewise, we propose a Feature Structuralized Domain Generalization (FSDG) model, wherein features experience structuralization into common, specific, and confounding segments, harmoniously aligned with their relevant semantic concepts, to elevate performance in FGDG. Specifically, feature structuralization (FS) is achieved through a decorrelation function on disentangled segments, constraints on common feature consistency, specific feature distinctiveness, and a prediction calibration operation across granularities. By imposing these stipulations, FSDG is prompted to disentangle and align features based on multi-granularity knowledge, facilitating robust subtle distinctions among categories. Extensive experimentation on three benchmarks consistently validates the superiority of FSDG over state-of-the-art counterparts, with an average improvement of 6.1% in terms of FGDG performance. Beyond that, the explainability analysis and experiments on various mainstream model architectures confirm the validity of FS.
Authors: Mehran Kazemi, Nishanth Dikkala, Ankit Anand, Petar Devic, Ishita Dasgupta, Fangyu Liu, Bahare Fatemi, Pranjal Awasthi, Dee Guo, Sreenivas Gollapudi, Ahmed Qureshi
Abstract: With the continuous advancement of large language models (LLMs), it is essential to create new benchmarks to effectively evaluate their expanding capabilities and identify areas for improvement. This work focuses on multi-image reasoning, an emerging capability in state-of-the-art LLMs. We introduce ReMI, a dataset designed to assess LLMs' ability to Reason with Multiple Images. This dataset encompasses a diverse range of tasks, spanning various reasoning domains such as math, physics, logic, code, table/chart understanding, and spatial and temporal reasoning. It also covers a broad spectrum of characteristics found in multi-image reasoning scenarios. We have benchmarked several cutting-edge LLMs using ReMI and found a substantial gap between their performance and human-level proficiency. This highlights the challenges in multi-image reasoning and the need for further research. Our analysis also reveals the strengths and weaknesses of different models, shedding light on the types of reasoning that are currently attainable and areas where future models require improvement. To foster further research in this area, we are releasing ReMI publicly: https://huggingface.co/datasets/mehrankazemi/ReMI.
Authors: Yijun Bei, Hengrui Lou, Jinsong Geng, Erteng Liu, Lechao Cheng, Jie Song, Mingli Song, Zunlei Feng
Abstract: With the rapid development of AI-generated content (AIGC) technology, the production of realistic fake facial images and videos that deceive human visual perception has become possible. Consequently, various face forgery detection techniques have been proposed to identify such fake facial content. However, evaluating the effectiveness and generalizability of these detection techniques remains a significant challenge. To address this, we have constructed a large-scale evaluation benchmark called DeepFaceGen, aimed at quantitatively assessing the effectiveness of face forgery detection and facilitating the iterative development of forgery detection technology. DeepFaceGen consists of 776,990 real face image/video samples and 773,812 face forgery image/video samples, generated using 34 mainstream face generation techniques. During the construction process, we carefully consider important factors such as content diversity, fairness across ethnicities, and availability of comprehensive labels, in order to ensure the versatility and convenience of DeepFaceGen. Subsequently, DeepFaceGen is employed in this study to evaluate and analyze the performance of 13 mainstream face forgery detection techniques from various perspectives. Through extensive experimental analysis, we derive significant findings and propose potential directions for future research. The code and dataset for DeepFaceGen are available at https://anonymous.4open.science/r/DeepFaceGen-47D1.
Authors: Syed Abdul Mateen, Niharika Malvia, Syed Abdul Khader, Danny Wang, Deepti Srinivasan, Chi-Fu Jeffrey Yang, Lana Schumacher, Sandeep Manjanna
Abstract: This paper presents an approach for surgical phase recognition using video data, aiming to provide a comprehensive understanding of surgical procedures for automated workflow analysis. The advent of robotic surgery, digitized operating rooms, and the generation of vast amounts of data have opened doors for the application of machine learning and computer vision in the analysis of surgical videos. Among these advancements, Surgical Phase Recognition(SPR) stands out as an emerging technology that has the potential to recognize and assess the ongoing surgical scenario, summarize the surgery, evaluate surgical skills, offer surgical decision support, and facilitate medical training. In this paper, we analyse and evaluate both frame-based and video clipping-based phase recognition on thoracic surgery dataset consisting of 11 classes of phases. Specifically, we utilize ImageNet ViT for image-based classification and VideoMAE as the baseline model for video-based classification. We show that Masked Video Distillation(MVD) exhibits superior performance, achieving a top-1 accuracy of 72.9%, compared to 52.31% achieved by ImageNet ViT. These findings underscore the efficacy of video-based classifiers over their image-based counterparts in surgical phase recognition tasks.
Authors: Jaeseok Byun, Seokhyeon Jeong, Wonjae Kim, Sanghyuk Chun, Taesup Moon
Abstract: Composed Image Retrieval (CIR) aims to retrieve a target image based on a reference image and conditioning text, enabling controllable searches. Due to the expensive dataset construction cost for CIR triplets, a zero-shot (ZS) CIR setting has been actively studied to eliminate the need for human-collected triplet datasets. The mainstream of ZS-CIR employs an efficient projection module that projects a CLIP image embedding to the CLIP text token embedding space, while fixing the CLIP encoders. Using the projected image embedding, these methods generate image-text composed features by using the pre-trained text encoder. However, their CLIP image and text encoders suffer from the task discrepancy between the pre-training task (text $\leftrightarrow$ image) and the target CIR task (image + text $\leftrightarrow$ image). Conceptually, we need expensive triplet samples to reduce the discrepancy, but we use cheap text triplets instead and update the text encoder. To that end, we introduce the Reducing Task Discrepancy of text encoders for Composed Image Retrieval (RTD), a plug-and-play training scheme for the text encoder that enhances its capability using a novel target-anchored text contrastive learning. We also propose two additional techniques to improve the proposed learning scheme: a hard negatives-based refined batch sampling strategy and a sophisticated concatenation scheme. Integrating RTD into the state-of-the-art projection-based ZS-CIR methods significantly improves performance across various datasets and backbones, demonstrating its efficiency and generalizability.
Authors: Ke Fan, Zechen Bai, Tianjun Xiao, Tong He, Max Horn, Yanwei Fu, Francesco Locatello, Zheng Zhang
Abstract: Object-centric learning (OCL) extracts the representation of objects with slots, offering an exceptional blend of flexibility and interpretability for abstracting low-level perceptual features. A widely adopted method within OCL is slot attention, which utilizes attention mechanisms to iteratively refine slot representations. However, a major drawback of most object-centric models, including slot attention, is their reliance on predefining the number of slots. This not only necessitates prior knowledge of the dataset but also overlooks the inherent variability in the number of objects present in each instance. To overcome this fundamental limitation, we present a novel complexity-aware object auto-encoder framework. Within this framework, we introduce an adaptive slot attention (AdaSlot) mechanism that dynamically determines the optimal number of slots based on the content of the data. This is achieved by proposing a discrete slot sampling module that is responsible for selecting an appropriate number of slots from a candidate list. Furthermore, we introduce a masked slot decoder that suppresses unselected slots during the decoding process. Our framework, tested extensively on object discovery tasks with various datasets, shows performance matching or exceeding top fixed-slot models. Moreover, our analysis substantiates that our method exhibits the capability to dynamically adapt the slot number according to each instance's complexity, offering the potential for further exploration in slot attention research. Project will be available at https://kfan21.github.io/AdaSlot/
Authors: Shuang Li, Jiaxu Leng, Guozhang Li, Ji Gan, Haosheng chen, Xinbo Gao
Abstract: Contrastive Language-Image Pre-Training (CLIP) has shown impressive performance in short-term Person Re-Identification (ReID) due to its ability to extract high-level semantic features of pedestrians, yet its direct application to Cloth-Changing Person Re-Identification (CC-ReID) faces challenges due to CLIP's image encoder overly focusing on clothes clues. To address this, we propose a novel framework called CLIP-Driven Cloth-Agnostic Feature Learning (CCAF) for CC-ReID. Accordingly, two modules were custom-designed: the Invariant Feature Prompting (IFP) and the Clothes Feature Minimization (CFM). These modules guide the model to extract cloth-agnostic features positively and attenuate clothes-related features negatively. Specifically, IFP is designed to extract fine-grained semantic features unrelated to clothes from the raw image, guided by the cloth-agnostic text prompts. This module first covers the clothes in the raw image at the pixel level to obtain the shielding image and then utilizes CLIP's knowledge to generate cloth-agnostic text prompts. Subsequently, it aligns the raw image-text and the raw image-shielding image in the feature space, emphasizing discriminative clues related to identity but unrelated to clothes. Furthermore, CFM is designed to examine and weaken the image encoder's ability to extract clothes features. It first generates text prompts corresponding to clothes pixels. Then, guided by these clothes text prompts, it iteratively examines and disentangles clothes features from pedestrian features, ultimately retaining inherent discriminative features. Extensive experiments have demonstrated the effectiveness of the proposed CCAF, achieving new state-of-the-art performance on several popular CC-ReID benchmarks without any additional inference time.
Authors: Peixi Wu, Bosong Chai, Xuan Nie, Longquan Yan, Zeyu Wang, Qifan Zhou, Boning Wang
Abstract: In this technical report, we present our findings from the research conducted on the Vast Vocabulary Visual Detection (V3Det) dataset for Supervised Vast Vocabulary Visual Detection task. How to deal with complex categories and detection boxes has become a difficulty in this track. The original supervised detector is not suitable for this task. We have designed a series of improvements, including adjustments to the network structure, changes to the loss function, and design of training strategies. Our model has shown improvement over the baseline and achieved excellent rankings on the Leaderboard for both the Vast Vocabulary Object Detection (Supervised) track and the Open Vocabulary Object Detection (OVD) track of the V3Det Challenge 2024.
Authors: Kaavya Rekanar, Martin Hayes, Ganesh Sistu, Ciaran Eising
Abstract: Visual Question Answering (VQA) models play a critical role in enhancing the perception capabilities of autonomous driving systems by allowing vehicles to analyze visual inputs alongside textual queries, fostering natural interaction and trust between the vehicle and its occupants or other road users. This study investigates the attention patterns of humans compared to a VQA model when answering driving-related questions, revealing disparities in the objects observed. We propose an approach integrating filters to optimize the model's attention mechanisms, prioritizing relevant objects and improving accuracy. Utilizing the LXMERT model for a case study, we compare attention patterns of the pre-trained and Filter Integrated models, alongside human answers using images from the NuImages dataset, gaining insights into feature prioritization. We evaluated the models using a Subjective scoring framework which shows that the integration of the feature encoder filter has enhanced the performance of the VQA model by refining its attention mechanisms.
Authors: Luk\'a\v{s} Adam, Vojt\v{e}ch \v{C}erm\'ak, Kostas Papafitsoros, Lukas Picek
Abstract: We introduce a new wildlife re-identification dataset WildlifeReID-10k with more than 214k images of 10k individual animals. It is a collection of 30 existing wildlife re-identification datasets with additional processing steps. WildlifeReID-10k contains animals as diverse as marine turtles, primates, birds, African herbivores, marine mammals and domestic animals. Due to the ubiquity of similar images in datasets, we argue that the standard (random) splits into training and testing sets are inadequate for wildlife re-identification and propose a new similarity-aware split based on the similarity of extracted features. To promote fair method comparison, we include similarity-aware splits both for closed-set and open-set settings, use MegaDescriptor - a foundational model for wildlife re-identification - for baseline performance and host a leaderboard with the best results. We publicly publish the dataset and the codes used to create it in the wildlife-datasets library, making WildlifeReID-10k both highly curated and easy to use.
Authors: Lianwei Yang, Zhikai Li, Junrui Xiao, Haisong Gong, Qingyi Gu
Abstract: Post-training quantization (PTQ) efficiently compresses vision models, but unfortunately, it accompanies a certain degree of accuracy degradation. Reconstruction methods aim to enhance model performance by narrowing the gap between the quantized model and the full-precision model, often yielding promising results. However, efforts to significantly improve the performance of PTQ through reconstruction in the Vision Transformer (ViT) have shown limited efficacy. In this paper, we conduct a thorough analysis of the reasons for this limited effectiveness and propose MGRQ (Mixed Granularity Reconstruction Quantization) as a solution to address this issue. Unlike previous reconstruction schemes, MGRQ introduces a mixed granularity reconstruction approach. Specifically, MGRQ enhances the performance of PTQ by introducing Extra-Block Global Supervision and Intra-Block Local Supervision, building upon Optimized Block-wise Reconstruction. Extra-Block Global Supervision considers the relationship between block outputs and the model's output, aiding block-wise reconstruction through global supervision. Meanwhile, Intra-Block Local Supervision reduces generalization errors by aligning the distribution of outputs at each layer within a block. Subsequently, MGRQ is further optimized for reconstruction through Mixed Granularity Loss Fusion. Extensive experiments conducted on various ViT models illustrate the effectiveness of MGRQ. Notably, MGRQ demonstrates robust performance in low-bit quantization, thereby enhancing the practicality of the quantized model.
Authors: Wei Lin, Muhammad Jehanzeb Mirza, Sivan Doveh, Rogerio Feris, Raja Giryes, Sepp Hochreiter, Leonid Karlinsky
Abstract: Comparing two images in terms of Commonalities and Differences (CaD) is a fundamental human capability that forms the basis of advanced visual reasoning and interpretation. It is essential for the generation of detailed and contextually relevant descriptions, performing comparative analysis, novelty detection, and making informed decisions based on visual data. However, surprisingly, little attention has been given to these fundamental concepts in the best current mimic of human visual intelligence - Large Multimodal Models (LMMs). We develop and contribute a new two-phase approach CaD-VI for collecting synthetic visual instructions, together with an instruction-following dataset CaD-Inst containing 349K image pairs with CaD instructions collected using CaD-VI. Our approach significantly improves the CaD spotting capabilities in LMMs, advancing the SOTA on a diverse set of related tasks by up to 17.5%. It is also complementary to existing difference-only instruction datasets, allowing automatic targeted refinement of those resources increasing their effectiveness for CaD tuning by up to 10%. Additionally, we propose an evaluation benchmark with 7.5K open-ended QAs to assess the CaD understanding abilities of LMMs.
Authors: Samar Fares, Klea Ziu, Toluwani Aremu, Nikita Durasov, Martin Tak\'a\v{c}, Pascal Fua, Karthik Nandakumar, Ivan Laptev
Abstract: Vision-Language Models (VLMs) are becoming increasingly vulnerable to adversarial attacks as various novel attack strategies are being proposed against these models. While existing defenses excel in unimodal contexts, they currently fall short in safeguarding VLMs against adversarial threats. To mitigate this vulnerability, we propose a novel, yet elegantly simple approach for detecting adversarial samples in VLMs. Our method leverages Text-to-Image (T2I) models to generate images based on captions produced by target VLMs. Subsequently, we calculate the similarities of the embeddings of both input and generated images in the feature space to identify adversarial samples. Empirical evaluations conducted on different datasets validate the efficacy of our approach, outperforming baseline methods adapted from image classification domains. Furthermore, we extend our methodology to classification tasks, showcasing its adaptability and model-agnostic nature. Theoretical analyses and empirical findings also show the resilience of our approach against adaptive attacks, positioning it as an excellent defense mechanism for real-world deployment against adversarial threats.
Authors: Maneesha Wickramasuriya, Taeyoung Lee, Murray Snyder
Abstract: This paper introduces a deep transformer network for estimating the relative 6D pose of a Unmanned Aerial Vehicle (UAV) with respect to a ship using monocular images. A synthetic dataset of ship images is created and annotated with 2D keypoints of multiple ship parts. A Transformer Neural Network model is trained to detect these keypoints and estimate the 6D pose of each part. The estimates are integrated using Bayesian fusion. The model is tested on synthetic data and in-situ flight experiments, demonstrating robustness and accuracy in various lighting conditions. The position estimation error is approximately 0.8\% and 1.0\% of the distance to the ship for the synthetic data and the flight experiments, respectively. The method has potential applications for ship-based autonomous UAV landing and navigation.
Authors: Changan Chen, Puyuan Peng, Ami Baid, Zihui Xue, Wei-Ning Hsu, David Harwarth, Kristen Grauman
Abstract: Generating realistic audio for human interactions is important for many applications, such as creating sound effects for films or virtual reality games. Existing approaches implicitly assume total correspondence between the video and audio during training, yet many sounds happen off-screen and have weak to no correspondence with the visuals -- resulting in uncontrolled ambient sounds or hallucinations at test time. We propose a novel ambient-aware audio generation model, AV-LDM. We devise a novel audio-conditioning mechanism to learn to disentangle foreground action sounds from the ambient background sounds in in-the-wild training videos. Given a novel silent video, our model uses retrieval-augmented generation to create audio that matches the visual content both semantically and temporally. We train and evaluate our model on two in-the-wild egocentric video datasets Ego4D and EPIC-KITCHENS. Our model outperforms an array of existing methods, allows controllable generation of the ambient sound, and even shows promise for generalizing to computer graphics game clips. Overall, our work is the first to focus video-to-audio generation faithfully on the observed visual content despite training from uncurated clips with natural background sounds.
Authors: Ziyi Wu, Yulia Rubanova, Rishabh Kabra, Drew A. Hudson, Igor Gilitschenski, Yusuf Aytar, Sjoerd van Steenkiste, Kelsey R. Allen, Thomas Kipf
Abstract: We address the problem of multi-object 3D pose control in image diffusion models. Instead of conditioning on a sequence of text tokens, we propose to use a set of per-object representations, Neural Assets, to control the 3D pose of individual objects in a scene. Neural Assets are obtained by pooling visual representations of objects from a reference image, such as a frame in a video, and are trained to reconstruct the respective objects in a different image, e.g., a later frame in the video. Importantly, we encode object visuals from the reference image while conditioning on object poses from the target frame. This enables learning disentangled appearance and pose features. Combining visual and 3D pose representations in a sequence-of-tokens format allows us to keep the text-to-image architecture of existing models, with Neural Assets in place of text tokens. By fine-tuning a pre-trained text-to-image diffusion model with this information, our approach enables fine-grained 3D pose and placement control of individual objects in a scene. We further demonstrate that Neural Assets can be transferred and recomposed across different scenes. Our model achieves state-of-the-art multi-object editing results on both synthetic 3D scene datasets, as well as two real-world video datasets (Objectron, Waymo Open).
Authors: Giuseppe Vecchio
Abstract: We introduce StableMaterials, a novel approach for generating photorealistic physical-based rendering (PBR) materials that integrate semi-supervised learning with Latent Diffusion Models (LDMs). Our method employs adversarial training to distill knowledge from existing large-scale image generation models, minimizing the reliance on annotated data and enhancing the diversity in generation. This distillation approach aligns the distribution of the generated materials with that of image textures from an SDXL model, enabling the generation of novel materials that are not present in the initial training dataset. Furthermore, we employ a diffusion-based refiner model to improve the visual quality of the samples and achieve high-resolution generation. Finally, we distill a latent consistency model for fast generation in just four steps and propose a new tileability technique that removes visual artifacts typically associated with fewer diffusion steps. We detail the architecture and training process of StableMaterials, the integration of semi-supervised training within existing LDM frameworks and show the advantages of our approach. Comparative evaluations with state-of-the-art methods show the effectiveness of StableMaterials, highlighting its potential applications in computer graphics and beyond. StableMaterials is publicly available at https://gvecchio.com/stablematerials.
Authors: Athmanarayanan Lakshmi Narayanan, Ranganath Krishnan, Amrutha Machireddy, Mahesh Subedar
Abstract: Foundational vision transformer models have shown impressive few shot performance on many vision tasks. This research presents a novel investigation into the application of parameter efficient fine-tuning methods within an active learning (AL) framework, to advance the sampling selection process in extremely budget constrained classification tasks. The focus on image datasets, known for their out-of-distribution characteristics, adds a layer of complexity and relevance to our study. Through a detailed evaluation, we illustrate the improved AL performance on these challenging datasets, highlighting the strategic advantage of merging parameter efficient fine tuning methods with foundation models. This contributes to the broader discourse on optimizing AL strategies, presenting a promising avenue for future exploration in leveraging foundation models for efficient and effective data annotation in specialized domains.
Authors: Yufan Zhou, Ruiyi Zhang, Kaizhi Zheng, Nanxuan Zhao, Jiuxiang Gu, Zichao Wang, Xin Eric Wang, Tong Sun
Abstract: In subject-driven text-to-image generation, recent works have achieved superior performance by training the model on synthetic datasets containing numerous image pairs. Trained on these datasets, generative models can produce text-aligned images for specific subject from arbitrary testing image in a zero-shot manner. They even outperform methods which require additional fine-tuning on testing images. However, the cost of creating such datasets is prohibitive for most researchers. To generate a single training pair, current methods fine-tune a pre-trained text-to-image model on the subject image to capture fine-grained details, then use the fine-tuned model to create images for the same subject based on creative text prompts. Consequently, constructing a large-scale dataset with millions of subjects can require hundreds of thousands of GPU hours. To tackle this problem, we propose Toffee, an efficient method to construct datasets for subject-driven editing and generation. Specifically, our dataset construction does not need any subject-level fine-tuning. After pre-training two generative models, we are able to generate infinite number of high-quality samples. We construct the first large-scale dataset for subject-driven image editing and generation, which contains 5 million image pairs, text prompts, and masks. Our dataset is 5 times the size of previous largest dataset, yet our cost is tens of thousands of GPU hours lower. To test the proposed dataset, we also propose a model which is capable of both subject-driven image editing and generation. By simply training the model on our proposed dataset, it obtains competitive results, illustrating the effectiveness of the proposed dataset construction framework.
Authors: Weiyi Wu, Chongyang Gao, Xinwen Xu, Siting Li, Jiang Gui
Abstract: Whole Slide Images (WSIs) are crucial for modern pathological diagnosis, yet their gigapixel-scale resolutions and sparse informative regions pose significant computational challenges. Traditional dense attention mechanisms, widely used in computer vision and natural language processing, are impractical for WSI analysis due to the substantial data scale and the redundant processing of uninformative areas. To address these challenges, we propose Memory-Efficient Sparse Pyramid Attention Networks with Shifted Windows (SPAN), drawing inspiration from state-of-the-art sparse attention techniques in other domains. SPAN introduces a sparse pyramid attention architecture that hierarchically focuses on informative regions within the WSI, aiming to reduce memory overhead while preserving critical features. Additionally, the incorporation of shifted windows enables the model to capture long-range contextual dependencies essential for accurate classification. We evaluated SPAN on multiple public WSI datasets, observing its competitive performance. Unlike existing methods that often struggle to model spatial and contextual information due to memory constraints, our approach enables the accurate modeling of these crucial features. Our study also highlights the importance of key design elements in attention mechanisms, such as the shifted-window scheme and the hierarchical structure, which contribute substantially to the effectiveness of SPAN in WSI analysis. The potential of SPAN for memory-efficient and effective analysis of WSI data is thus demonstrated, and the code will be made publicly available following the publication of this work.
Authors: Chunyi Li, Xiele Wu, Haoning Wu, Donghui Feng, Zicheng Zhang, Guo Lu, Xiongkuo Min, Xiaohong Liu, Guangtao Zhai, Weisi Lin
Abstract: Ultra-low bitrate image compression is a challenging and demanding topic. With the development of Large Multimodal Models (LMMs), a Cross Modality Compression (CMC) paradigm of Image-Text-Image has emerged. Compared with traditional codecs, this semantic-level compression can reduce image data size to 0.1\% or even lower, which has strong potential applications. However, CMC has certain defects in consistency with the original image and perceptual quality. To address this problem, we introduce CMC-Bench, a benchmark of the cooperative performance of Image-to-Text (I2T) and Text-to-Image (T2I) models for image compression. This benchmark covers 18,000 and 40,000 images respectively to verify 6 mainstream I2T and 12 T2I models, including 160,000 subjective preference scores annotated by human experts. At ultra-low bitrates, this paper proves that the combination of some I2T and T2I models has surpassed the most advanced visual signal codecs; meanwhile, it highlights where LMMs can be further optimized toward the compression task. We encourage LMM developers to participate in this test to promote the evolution of visual signal codec protocols.
Authors: Zijia Zhao, Haoyu Lu, Yuqi Huo, Yifan Du, Tongtian Yue, Longteng Guo, Bingning Wang, Weipeng Chen, Jing Liu
Abstract: Video understanding is a crucial next step for multimodal large language models (MLLMs). To probe specific aspects of video understanding ability, existing video benchmarks typically require careful video selection based on the target capability, along with laborious annotation of query-response pairs to match the specific video content. This process is both challenging and resource-intensive. In this paper, we propose VideoNIAH (Video Needle In A Haystack), a benchmark construction framework through synthetic video generation. VideoNIAH decouples test video content from their query-responses by inserting unrelated image/text 'needles' into original videos. It generates annotations solely from these needles, ensuring diversity in video sources and a variety of query-responses. Additionally, by inserting multiple needles, VideoNIAH rigorously evaluates the temporal understanding capabilities of models. We utilized VideoNIAH to compile a video benchmark VNBench, including tasks such as retrieval, ordering, and counting. VNBench can efficiently evaluate the fine-grained understanding ability and spatio-temporal modeling ability of a video model, while also supporting the long-context evaluation. Additionally, we evaluated recent video-centric multimodal large language models (MLLMs), both open-source and proprietary, providing a comprehensive analysis. We found that although proprietary models have significant advantages over open-source models, all existing video models still perform poorly on long-distance dependency tasks. VideoNIAH is a simple yet highly scalable benchmark construction framework, and we believe it will inspire future video benchmark works. The code and data are available at https://github.com/joez17/VideoNIAH.
Authors: Yigit Ekin, Ahmet Burak Yildirim, Erdem Eren Caglar, Aykut Erdem, Erkut Erdem, Aysegul Dundar
Abstract: Advanced image editing techniques, particularly inpainting, are essential for seamlessly removing unwanted elements while preserving visual integrity. Traditional GAN-based methods have achieved notable success, but recent advancements in diffusion models have produced superior results due to their training on large-scale datasets, enabling the generation of remarkably realistic inpainted images. Despite their strengths, diffusion models often struggle with object removal tasks without explicit guidance, leading to unintended hallucinations of the removed object. To address this issue, we introduce CLIPAway, a novel approach leveraging CLIP embeddings to focus on background regions while excluding foreground elements. CLIPAway enhances inpainting accuracy and quality by identifying embeddings that prioritize the background, thus achieving seamless object removal. Unlike other methods that rely on specialized training datasets or costly manual annotations, CLIPAway provides a flexible, plug-and-play solution compatible with various diffusion-based inpainting techniques.
Authors: Desai Xie, Sai Bi, Zhixin Shu, Kai Zhang, Zexiang Xu, Yi Zhou, S\"oren Pirk, Arie Kaufman, Xin Sun, Hao Tan
Abstract: We present LRM-Zero, a Large Reconstruction Model (LRM) trained entirely on synthesized 3D data, achieving high-quality sparse-view 3D reconstruction. The core of LRM-Zero is our procedural 3D dataset, Zeroverse, which is automatically synthesized from simple primitive shapes with random texturing and augmentations (e.g., height fields, boolean differences, and wireframes). Unlike previous 3D datasets (e.g., Objaverse) which are often captured or crafted by humans to approximate real 3D data, Zeroverse completely ignores realistic global semantics but is rich in complex geometric and texture details that are locally similar to or even more intricate than real objects. We demonstrate that our LRM-Zero, trained with our fully synthesized Zeroverse, can achieve high visual quality in the reconstruction of real-world objects, competitive with models trained on Objaverse. We also analyze several critical design choices of Zeroverse that contribute to LRM-Zero's capability and training stability. Our work demonstrates that 3D reconstruction, one of the core tasks in 3D vision, can potentially be addressed without the semantics of real-world objects. The Zeroverse's procedural synthesis code and interactive visualization are available at: https://desaixie.github.io/lrm-zero/.
Authors: S. Mahdi H. Miangoleh, Mahesh Reddy, Ya\u{g}{\i}z Aksoy
Abstract: Existing methods for scale-invariant monocular depth estimation (SI MDE) often struggle due to the complexity of the task, and limited and non-diverse datasets, hindering generalizability in real-world scenarios. This is while shift-and-scale-invariant (SSI) depth estimation, simplifying the task and enabling training with abundant stereo datasets achieves high performance. We present a novel approach that leverages SSI inputs to enhance SI depth estimation, streamlining the network's role and facilitating in-the-wild generalization for SI depth estimation while only using a synthetic dataset for training. Emphasizing the generation of high-resolution details, we introduce a novel sparse ordinal loss that substantially improves detail generation in SSI MDE, addressing critical limitations in existing approaches. Through in-the-wild qualitative examples and zero-shot evaluation we substantiate the practical utility of our approach in computational photography applications, showcasing its ability to generate highly detailed SI depth maps and achieve generalization in diverse scenarios.
Authors: Tobias Kirschstein, Simon Giebenhain, Jiapeng Tang, Markos Georgopoulos, Matthias Nie{\ss}ner
Abstract: Learning 3D head priors from large 2D image collections is an important step towards high-quality 3D-aware human modeling. A core requirement is an efficient architecture that scales well to large-scale datasets and large image resolutions. Unfortunately, existing 3D GANs struggle to scale to generate samples at high resolutions due to their relatively slow train and render speeds, and typically have to rely on 2D superresolution networks at the expense of global 3D consistency. To address these challenges, we propose Generative Gaussian Heads (GGHead), which adopts the recent 3D Gaussian Splatting representation within a 3D GAN framework. To generate a 3D representation, we employ a powerful 2D CNN generator to predict Gaussian attributes in the UV space of a template head mesh. This way, GGHead exploits the regularity of the template's UV layout, substantially facilitating the challenging task of predicting an unstructured set of 3D Gaussians. We further improve the geometric fidelity of the generated 3D representations with a novel total variation loss on rendered UV coordinates. Intuitively, this regularization encourages that neighboring rendered pixels should stem from neighboring Gaussians in the template's UV space. Taken together, our pipeline can efficiently generate 3D heads trained only from single-view 2D image observations. Our proposed framework matches the quality of existing 3D head GANs on FFHQ while being both substantially faster and fully 3D consistent. As a result, we demonstrate real-time generation and rendering of high-quality 3D-consistent heads at $1024^2$ resolution for the first time.
Authors: Yiming Li, Zhiheng Li, Nuo Chen, Moonjun Gong, Zonglin Lyu, Zehong Wang, Peili Jiang, Chen Feng
Abstract: Large-scale datasets have fueled recent advancements in AI-based autonomous vehicle research. However, these datasets are usually collected from a single vehicle's one-time pass of a certain location, lacking multiagent interactions or repeated traversals of the same place. Such information could lead to transformative enhancements in autonomous vehicles' perception, prediction, and planning capabilities. To bridge this gap, in collaboration with the self-driving company May Mobility, we present the MARS dataset which unifies scenarios that enable MultiAgent, multitraveRSal, and multimodal autonomous vehicle research. More specifically, MARS is collected with a fleet of autonomous vehicles driving within a certain geographical area. Each vehicle has its own route and different vehicles may appear at nearby locations. Each vehicle is equipped with a LiDAR and surround-view RGB cameras. We curate two subsets in MARS: one facilitates collaborative driving with multiple vehicles simultaneously present at the same location, and the other enables memory retrospection through asynchronous traversals of the same location by multiple vehicles. We conduct experiments in place recognition and neural reconstruction. More importantly, MARS introduces new research opportunities and challenges such as multitraversal 3D reconstruction, multiagent perception, and unsupervised object discovery. Our data and codes can be found at https://ai4ce.github.io/MARS/.
Authors: Yue Zhou, Litong Feng, Yiping Ke, Xue Jiang, Junchi Yan, Xue Yang, Wayne Zhang
Abstract: Vision-Language Foundation Models (VLFMs) have made remarkable progress on various multimodal tasks, such as image captioning, image-text retrieval, visual question answering, and visual grounding. However, most methods rely on training with general image datasets, and the lack of geospatial data leads to poor performance on earth observation. Numerous geospatial image-text pair datasets and VLFMs fine-tuned on them have been proposed recently. These new approaches aim to leverage large-scale, multimodal geospatial data to build versatile intelligent models with diverse geo-perceptive capabilities, which we refer to as Vision-Language Geo-Foundation Models (VLGFMs). This paper thoroughly reviews VLGFMs, summarizing and analyzing recent developments in the field. In particular, we introduce the background and motivation behind the rise of VLGFMs, highlighting their unique research significance. Then, we systematically summarize the core technologies employed in VLGFMs, including data construction, model architectures, and applications of various multimodal geospatial tasks. Finally, we conclude with insights, issues, and discussions regarding future research directions. To the best of our knowledge, this is the first comprehensive literature review of VLGFMs. We keep tracing related works at https://github.com/zytx121/Awesome-VLGFM.
Authors: Yunsong Zhou, Michael Simon, Zhenghao Peng, Sicheng Mo, Hongzi Zhu, Minyi Guo, Bolei Zhou
Abstract: Controllable synthetic data generation can substantially lower the annotation cost of training data in autonomous driving research and development. Prior works use diffusion models to generate driving images conditioned on the 3D object layout. However, those models are trained on small-scale datasets like nuScenes, which lack appearance and layout diversity. Moreover, the trained models can only generate images based on the real-world layout data from the validation set of the same dataset, where overfitting might happen. In this work, we introduce a simulator-conditioned scene generation framework called SimGen that can learn to generate diverse driving scenes by mixing data from the simulator and the real world. It uses a novel cascade diffusion pipeline to address challenging sim-to-real gaps and multi-condition conflicts. A driving video dataset DIVA is collected to enhance the generative diversity of SimGen, which contains over 147.5 hours of real-world driving videos from 73 locations worldwide and simulated driving data from the MetaDrive simulator. SimGen achieves superior generation quality and diversity while preserving controllability based on the text prompt and the layout pulled from a simulator. We further demonstrate the improvements brought by SimGen for synthetic data augmentation on the BEV detection and segmentation task and showcase its capability in safety-critical data generation. Code, data, and models will be made available.
Authors: Youngtaek Oh, Pyunghwan Ahn, Jinhyung Kim, Gwangmo Song, Soonyoung Lee, In So Kweon, Junmo Kim
Abstract: Vision and language models (VLMs) such as CLIP have showcased remarkable zero-shot recognition abilities yet face challenges in visio-linguistic compositionality, particularly in linguistic comprehension and fine-grained image-text alignment. This paper explores the intricate relationship between compositionality and recognition -- two pivotal aspects of VLM capability. We conduct a comprehensive evaluation of existing VLMs, covering both pre-training approaches aimed at recognition and the fine-tuning methods designed to improve compositionality. Our evaluation employs 12 benchmarks for compositionality, along with 21 zero-shot classification and two retrieval benchmarks for recognition. In our analysis from 274 CLIP model checkpoints, we reveal patterns and trade-offs that emerge between compositional understanding and recognition accuracy. Ultimately, this necessitates strategic efforts towards developing models that improve both capabilities, as well as the meticulous formulation of benchmarks for compositionality. We open our evaluation framework at https://github.com/ytaek-oh/vl_compo.
Authors: Rajatsubhra Chakraborty, Arkaprava Sinha, Dominick Reilly, Manish Kumar Govind, Pu Wang, Francois Bremond, Srijan Das
Abstract: Large Language Vision Models (LLVMs) have demonstrated effectiveness in processing internet videos, yet they struggle with the visually perplexing dynamics present in Activities of Daily Living (ADL) due to limited pertinent datasets and models tailored to relevant cues. To this end, we propose a framework for curating ADL multiview datasets to fine-tune LLVMs, resulting in the creation of ADL-X, comprising 100K RGB video-instruction pairs, language descriptions, 3D skeletons, and action-conditioned object trajectories. We introduce LLAVIDAL, an LLVM capable of incorporating 3D poses and relevant object trajectories to understand the intricate spatiotemporal relationships within ADLs. Furthermore, we present a novel benchmark, ADLMCQ, for quantifying LLVM effectiveness in ADL scenarios. When trained on ADL-X, LLAVIDAL consistently achieves state-of-the-art performance across all ADL evaluation metrics. Qualitative analysis reveals LLAVIDAL's temporal reasoning capabilities in understanding ADL. The link to the dataset is provided at: https://adl-x.github.io/
URLs: https://adl-x.github.io/
Authors: Hong-Xing Yu, Haoyi Duan, Charles Herrmann, William T. Freeman, Jiajun Wu
Abstract: We present WonderWorld, a novel framework for \emph{interactive} 3D scene extrapolation that enables users to explore and shape virtual environments based on a single input image and user-specified text. While significant improvements have been made to the visual quality of scene generation, existing methods are run offline, taking tens of minutes to hours to generate a scene. By leveraging Fast Gaussian Surfels and a guided diffusion-based depth estimation method, WonderWorld generates geometrically consistent extrapolation while significantly reducing computational time. Our framework generates connected and diverse 3D scenes in less than 10 seconds on a single A6000 GPU, enabling real-time user interaction and exploration. We demonstrate the potential of WonderWorld for applications in virtual reality, gaming, and creative design, where users can quickly generate and navigate immersive, potentially infinite virtual worlds from a single image. Our approach represents a significant advancement in interactive 3D scene generation, opening up new possibilities for user-driven content creation and exploration in virtual environments. We will release full code and software for reproducibility. Project website: https://WonderWorld-2024.github.io/
Authors: Meng-Li Shih, Jia-Bin Huang, Changil Kim, Rajvi Shah, Johannes Kopf, Chen Gao
Abstract: We introduce a novel method for dynamic free-view synthesis of an ambient scenes from a monocular capture bringing a immersive quality to the viewing experience. Our method builds upon the recent advancements in 3D Gaussian Splatting (3DGS) that can faithfully reconstruct complex static scenes. Previous attempts to extend 3DGS to represent dynamics have been confined to bounded scenes or require multi-camera captures, and often fail to generalize to unseen motions, limiting their practical application. Our approach overcomes these constraints by leveraging the periodicity of ambient motions to learn the motion trajectory model, coupled with careful regularization. We also propose important practical strategies to improve the visual quality of the baseline 3DGS static reconstructions and to improve memory efficiency critical for GPU-memory intensive learning. We demonstrate high-quality photorealistic novel view synthesis of several ambient natural scenes with intricate textures and fine structural elements.
Authors: Jongwoo Park, Kanchana Ranasinghe, Kumara Kahatapitiya, Wonjeong Ryoo, Donghyun Kim, Michael S. Ryoo
Abstract: Long-form videos that span across wide temporal intervals are highly information redundant and contain multiple distinct events or entities that are often loosely-related. Therefore, when performing long-form video question answering (LVQA),all information necessary to generate a correct response can often be contained within a small subset of frames. Recent literature explore the use of large language models (LLMs) in LVQA benchmarks, achieving exceptional performance, while relying on vision language models (VLMs) to convert all visual content within videos into natural language. Such VLMs often independently caption a large number of frames uniformly sampled from long videos, which is not efficient and can mostly be redundant. Questioning these decision choices, we explore optimal strategies for key-frame selection and sequence-aware captioning, that can significantly reduce these redundancies. We propose two novel approaches that improve each of aspects, namely Hierarchical Keyframe Selector and Sequential Visual LLM. Our resulting framework termed LVNet achieves state-of-the-art performance across three benchmark LVQA datasets. Our code will be released publicly.
Authors: Miaosen Zhang, Yixuan Wei, Zhen Xing, Yifei Ma, Zuxuan Wu, Ji Li, Zheng Zhang, Qi Dai, Chong Luo, Xin Geng, Baining Guo
Abstract: Modern vision models are trained on very large noisy datasets. While these models acquire strong capabilities, they may not follow the user's intent to output the desired results in certain aspects, e.g., visual aesthetic, preferred style, and responsibility. In this paper, we target the realm of visual aesthetics and aim to align vision models with human aesthetic standards in a retrieval system. Advanced retrieval systems usually adopt a cascade of aesthetic models as re-rankers or filters, which are limited to low-level features like saturation and perform poorly when stylistic, cultural or knowledge contexts are involved. We find that utilizing the reasoning ability of large language models (LLMs) to rephrase the search query and extend the aesthetic expectations can make up for this shortcoming. Based on the above findings, we propose a preference-based reinforcement learning method that fine-tunes the vision models to distill the knowledge from both LLMs reasoning and the aesthetic models to better align the vision models with human aesthetics. Meanwhile, with rare benchmarks designed for evaluating retrieval systems, we leverage large multi-modality model (LMM) to evaluate the aesthetic performance with their strong abilities. As aesthetic assessment is one of the most subjective tasks, to validate the robustness of LMM, we further propose a novel dataset named HPIR to benchmark the alignment with human aesthetics. Experiments demonstrate that our method significantly enhances the aesthetic behaviors of the vision models, under several metrics. We believe the proposed algorithm can be a general practice for aligning vision models with human values.
Authors: Bar Cavia, Eliahu Horwitz, Tal Reiss, Yedid Hoshen
Abstract: Recent improvements in generative AI made synthesizing fake images easy; as they can be used to cause harm, it is crucial to develop accurate techniques to identify them. This paper introduces "Locally Aware Deepfake Detection Algorithm" (LaDeDa), that accepts a single 9x9 image patch and outputs its deepfake score. The image deepfake score is the pooled score of its patches. With merely patch-level information, LaDeDa significantly improves over the state-of-the-art, achieving around 99% mAP on current benchmarks. Owing to the patch-level structure of LaDeDa, we hypothesize that the generation artifacts can be detected by a simple model. We therefore distill LaDeDa into Tiny-LaDeDa, a highly efficient model consisting of only 4 convolutional layers. Remarkably, Tiny-LaDeDa has 375x fewer FLOPs and is 10,000x more parameter-efficient than LaDeDa, allowing it to run efficiently on edge devices with a minor decrease in accuracy. These almost-perfect scores raise the question: is the task of deepfake detection close to being solved? Perhaps surprisingly, our investigation reveals that current training protocols prevent methods from generalizing to real-world deepfakes extracted from social media. To address this issue, we introduce WildRF, a new deepfake detection dataset curated from several popular social networks. Our method achieves the top performance of 93.7% mAP on WildRF, however the large gap from perfect accuracy shows that reliable real-world deepfake detection is still unsolved.
Authors: Junke Wang, Yi Jiang, Zehuan Yuan, Binyue Peng, Zuxuan Wu, Yu-Gang Jiang
Abstract: Tokenizer, serving as a translator to map the intricate visual data into a compact latent space, lies at the core of visual generative models. Based on the finding that existing tokenizers are tailored to image or video inputs, this paper presents OmniTokenizer, a transformer-based tokenizer for joint image and video tokenization. OmniTokenizer is designed with a spatial-temporal decoupled architecture, which integrates window and causal attention for spatial and temporal modeling. To exploit the complementary nature of image and video data, we further propose a progressive training strategy, where OmniTokenizer is first trained on image data on a fixed resolution to develop the spatial encoding capacity and then jointly trained on image and video data on multiple resolutions to learn the temporal dynamics. OmniTokenizer, for the first time, handles both image and video inputs within a unified framework and proves the possibility of realizing their synergy. Extensive experiments demonstrate that OmniTokenizer achieves state-of-the-art (SOTA) reconstruction performance on various image and video datasets, e.g., 1.11 reconstruction FID on ImageNet and 42 reconstruction FVD on UCF-101, beating the previous SOTA methods by 13% and 26%, respectively. Additionally, we also show that when integrated with OmniTokenizer, both language model-based approaches and diffusion models can realize advanced visual synthesis performance, underscoring the superiority and versatility of our method. Code is available at https://github.com/FoundationVision/OmniTokenizer.
Authors: Thao Nguyen, Haotian Liu, Yuheng Li, Mu Cai, Utkarsh Ojha, Yong Jae Lee
Abstract: Large Multimodal Models (LMMs) have shown remarkable capabilities across a variety of tasks (e.g., image captioning, visual question answering). While broad, their knowledge remains generic (e.g., recognizing a dog), and they are unable to handle personalized subjects (e.g., recognizing a user's pet dog). Human reasoning, in contrast, typically operates within the context of specific subjects in our surroundings. For example, one might ask, "What should I buy for my dog's birthday?"; as opposed to a generic inquiry about "What should I buy for a dog's birthday?". Similarly, when looking at a friend's image, the interest lies in seeing their activities (e.g., "my friend is holding a cat"), rather than merely observing generic human actions (e.g., "a man is holding a cat"). In this paper, we introduce the novel task of personalizing LMMs, so that they can have conversations about a specific subject. We propose Yo'LLaVA, which learns to embed a personalized subject into a set of latent tokens given a handful of example images of the subject. Our qualitative and quantitative analyses reveal that Yo'LLaVA can learn the concept more efficiently using fewer tokens and more effectively encode the visual attributes compared to strong prompting baselines (e.g., LLaVA).
Authors: Ruiyuan Lyu, Tai Wang, Jingli Lin, Shuai Yang, Xiaohan Mao, Yilun Chen, Runsen Xu, Haifeng Huang, Chenming Zhu, Dahua Lin, Jiangmiao Pang
Abstract: With the emergence of LLMs and their integration with other data modalities, multi-modal 3D perception attracts more attention due to its connectivity to the physical world and makes rapid progress. However, limited by existing datasets, previous works mainly focus on understanding object properties or inter-object spatial relationships in a 3D scene. To tackle this problem, this paper builds the first largest ever multi-modal 3D scene dataset and benchmark with hierarchical grounded language annotations, MMScan. It is constructed based on a top-down logic, from region to object level, from a single target to inter-target relationships, covering holistic aspects of spatial and attribute understanding. The overall pipeline incorporates powerful VLMs via carefully designed prompts to initialize the annotations efficiently and further involve humans' correction in the loop to ensure the annotations are natural, correct, and comprehensive. Built upon existing 3D scanning data, the resulting multi-modal 3D dataset encompasses 1.4M meta-annotated captions on 109k objects and 7.7k regions as well as over 3.04M diverse samples for 3D visual grounding and question-answering benchmarks. We evaluate representative baselines on our benchmarks, analyze their capabilities in different aspects, and showcase the key problems to be addressed in the future. Furthermore, we use this high-quality dataset to train state-of-the-art 3D visual grounding and LLMs and obtain remarkable performance improvement both on existing benchmarks and in-the-wild evaluation. Codes, datasets, and benchmarks will be available at https://github.com/OpenRobotLab/EmbodiedScan.
Authors: Linzhan Mou, Jun-Kun Chen, Yu-Xiong Wang
Abstract: This paper proposes Instruct 4D-to-4D that achieves 4D awareness and spatial-temporal consistency for 2D diffusion models to generate high-quality instruction-guided dynamic scene editing results. Traditional applications of 2D diffusion models in dynamic scene editing often result in inconsistency, primarily due to their inherent frame-by-frame editing methodology. Addressing the complexities of extending instruction-guided editing to 4D, our key insight is to treat a 4D scene as a pseudo-3D scene, decoupled into two sub-problems: achieving temporal consistency in video editing and applying these edits to the pseudo-3D scene. Following this, we first enhance the Instruct-Pix2Pix (IP2P) model with an anchor-aware attention module for batch processing and consistent editing. Additionally, we integrate optical flow-guided appearance propagation in a sliding window fashion for more precise frame-to-frame editing and incorporate depth-based projection to manage the extensive data of pseudo-3D scenes, followed by iterative editing to achieve convergence. We extensively evaluate our approach in various scenes and editing instructions, and demonstrate that it achieves spatially and temporally consistent editing results, with significantly enhanced detail and sharpness over the prior art. Notably, Instruct 4D-to-4D is general and applicable to both monocular and challenging multi-camera scenes. Code and more results are available at immortalco.github.io/Instruct-4D-to-4D.
Authors: Yushi Hu, Weijia Shi, Xingyu Fu, Dan Roth, Mari Ostendorf, Luke Zettlemoyer, Noah A Smith, Ranjay Krishna
Abstract: Humans draw to facilitate reasoning: we draw auxiliary lines when solving geometry problems; we mark and circle when reasoning on maps; we use sketches to amplify our ideas and relieve our limited-capacity working memory. However, such actions are missing in current multimodal language models (LMs). Current chain-of-thought and tool-use paradigms only use text as intermediate reasoning steps. In this work, we introduce Sketchpad, a framework that gives multimodal LMs a visual sketchpad and tools to draw on the sketchpad. The LM conducts planning and reasoning according to the visual artifacts it has drawn. Different from prior work, which uses text-to-image models to enable LMs to draw, Sketchpad enables LMs to draw with lines, boxes, marks, etc., which is closer to human sketching and better facilitates reasoning. Sketchpad can also use specialist vision models during the sketching process (e.g., draw bounding boxes with object detection models, draw masks with segmentation models), to further enhance visual perception and reasoning. We experiment with a wide range of math tasks (including geometry, functions, graphs, and chess) and complex visual reasoning tasks. Sketchpad substantially improves performance on all tasks over strong base models with no sketching, yielding an average gain of 12.7% on math tasks, and 8.6% on vision tasks. GPT-4o with Sketchpad sets a new state of the art on all tasks, including V*Bench (80.3%), BLINK spatial reasoning (83.9%), and visual correspondence (80.8%). All codes and data are in https://visualsketchpad.github.io/.
Authors: Jun-Kun Chen, Samuel Rota Bul\`o, Norman M\"uller, Lorenzo Porzi, Peter Kontschieder, Yu-Xiong Wang
Abstract: This paper proposes ConsistDreamer - a novel framework that lifts 2D diffusion models with 3D awareness and 3D consistency, thus enabling high-fidelity instruction-guided scene editing. To overcome the fundamental limitation of missing 3D consistency in 2D diffusion models, our key insight is to introduce three synergetic strategies that augment the input of the 2D diffusion model to become 3D-aware and to explicitly enforce 3D consistency during the training process. Specifically, we design surrounding views as context-rich input for the 2D diffusion model, and generate 3D-consistent, structured noise instead of image-independent noise. Moreover, we introduce self-supervised consistency-enforcing training within the per-scene editing procedure. Extensive evaluation shows that our ConsistDreamer achieves state-of-the-art performance for instruction-guided scene editing across various scenes and editing instructions, particularly in complicated large-scale indoor scenes from ScanNet++, with significantly improved sharpness and fine-grained textures. Notably, ConsistDreamer stands as the first work capable of successfully editing complex (e.g., plaid/checkered) patterns. Our project page is at immortalco.github.io/ConsistDreamer.
Authors: Roman Bachmann, O\u{g}uzhan Fatih Kar, David Mizrahi, Ali Garjani, Mingfei Gao, David Griffiths, Jiaming Hu, Afshin Dehghan, Amir Zamir
Abstract: Current multimodal and multitask foundation models like 4M or UnifiedIO show promising results, but in practice their out-of-the-box abilities to accept diverse inputs and perform diverse tasks are limited by the (usually rather small) number of modalities and tasks they are trained on. In this paper, we expand upon the capabilities of them by training a single model on tens of highly diverse modalities and by performing co-training on large-scale multimodal datasets and text corpora. This includes training on several semantic and geometric modalities, feature maps from recent state of the art models like DINOv2 and ImageBind, pseudo labels of specialist models like SAM and 4DHumans, and a range of new modalities that allow for novel ways to interact with the model and steer the generation, for example image metadata or color palettes. A crucial step in this process is performing discrete tokenization on various modalities, whether they are image-like, neural network feature maps, vectors, structured data like instance segmentation or human poses, or data that can be represented as text. Through this, we expand on the out-of-the-box capabilities of multimodal models and specifically show the possibility of training one model to solve at least 3x more tasks/modalities than existing ones and doing so without a loss in performance. This enables more fine-grained and controllable multimodal generation capabilities and allows us to study the distillation of models trained on diverse data and objectives into a unified model. We successfully scale the training to a three billion parameter model using tens of modalities and different datasets. The resulting models and training code are open sourced at 4m.epfl.ch.
Authors: Hashmat Shadab Malik, Fahad Shamshad, Muzammal Naseer, Karthik Nandakumar, Fahad Shahbaz Khan, Salman Khan
Abstract: Vision State Space Models (VSSMs), a novel architecture that combines the strengths of recurrent neural networks and latent variable models, have demonstrated remarkable performance in visual perception tasks by efficiently capturing long-range dependencies and modeling complex visual dynamics. However, their robustness under natural and adversarial perturbations remains a critical concern. In this work, we present a comprehensive evaluation of VSSMs' robustness under various perturbation scenarios, including occlusions, image structure, common corruptions, and adversarial attacks, and compare their performance to well-established architectures such as transformers and Convolutional Neural Networks. Furthermore, we investigate the resilience of VSSMs to object-background compositional changes on sophisticated benchmarks designed to test model performance in complex visual scenes. We also assess their robustness on object detection and segmentation tasks using corrupted datasets that mimic real-world scenarios. To gain a deeper understanding of VSSMs' adversarial robustness, we conduct a frequency analysis of adversarial attacks, evaluating their performance against low-frequency and high-frequency perturbations. Our findings highlight the strengths and limitations of VSSMs in handling complex visual corruptions, offering valuable insights for future research and improvements in this promising field. Our code and models will be available at https://github.com/HashmatShadab/MambaRobustness.
Authors: Sheng-Yu Wang, Aaron Hertzmann, Alexei A. Efros, Jun-Yan Zhu, Richard Zhang
Abstract: The goal of data attribution for text-to-image models is to identify the training images that most influence the generation of a new image. We can define "influence" by saying that, for a given output, if a model is retrained from scratch without that output's most influential images, the model should then fail to generate that output image. Unfortunately, directly searching for these influential images is computationally infeasible, since it would require repeatedly retraining from scratch. We propose a new approach that efficiently identifies highly-influential images. Specifically, we simulate unlearning the synthesized image, proposing a method to increase the training loss on the output image, without catastrophic forgetting of other, unrelated concepts. Then, we find training images that are forgotten by proxy, identifying ones with significant loss deviations after the unlearning process, and label these as influential. We evaluate our method with a computationally intensive but "gold-standard" retraining from scratch and demonstrate our method's advantages over previous methods.
Authors: Sachin Shah, Matthew Albert Chan, Haoming Cai, Jingxi Chen, Sakshum Kulshrestha, Chahat Deep Singh, Yiannis Aloimonos, Christopher Metzler
Abstract: Point-spread-function (PSF) engineering is a well-established computational imaging technique that uses phase masks and other optical elements to embed extra information (e.g., depth) into the images captured by conventional CMOS image sensors. To date, however, PSF-engineering has not been applied to neuromorphic event cameras; a powerful new image sensing technology that responds to changes in the log-intensity of light. This paper establishes theoretical limits (Cram\'er Rao bounds) on 3D point localization and tracking with PSF-engineered event cameras. Using these bounds, we first demonstrate that existing Fisher phase masks are already near-optimal for localizing static flashing point sources (e.g., blinking fluorescent molecules). We then demonstrate that existing designs are sub-optimal for tracking moving point sources and proceed to use our theory to design optimal phase masks and binary amplitude masks for this task. To overcome the non-convexity of the design problem, we leverage novel implicit neural representation based parameterizations of the phase and amplitude masks. We demonstrate the efficacy of our designs through extensive simulations. We also validate our method with a simple prototype.
Authors: Yansheng Li, Linlin Wang, Tingzhu Wang, Xue Yang, Junwei Luo, Qi Wang, Youming Deng, Wenbin Wang, Xian Sun, Haifeng Li, Bo Dang, Yongjun Zhang, Yi Yu, Junchi Yan
Abstract: Scene graph generation (SGG) in satellite imagery (SAI) benefits promoting intelligent understanding of geospatial scenarios from perception to cognition. In SAI, objects exhibit great variations in scales and aspect ratios, and there exist rich relationships between objects (even between spatially disjoint objects), which makes it necessary to holistically conduct SGG in large-size very-high-resolution (VHR) SAI. However, the lack of SGG datasets with large-size VHR SAI has constrained the advancement of SGG in SAI. Due to the complexity of large-size VHR SAI, mining triplets
Authors: Fei Wang, Xingyu Fu, James Y. Huang, Zekun Li, Qin Liu, Xiaogeng Liu, Mingyu Derek Ma, Nan Xu, Wenxuan Zhou, Kai Zhang, Tianyi Lorena Yan, Wenjie Jacky Mo, Hsiang-Hui Liu, Pan Lu, Chunyuan Li, Chaowei Xiao, Kai-Wei Chang, Dan Roth, Sheng Zhang, Hoifung Poon, Muhao Chen
Abstract: We introduce MuirBench, a comprehensive benchmark that focuses on robust multi-image understanding capabilities of multimodal LLMs. MuirBench consists of 12 diverse multi-image tasks (e.g., scene understanding, ordering) that involve 10 categories of multi-image relations (e.g., multiview, temporal relations). Comprising 11,264 images and 2,600 multiple-choice questions, MuirBench is created in a pairwise manner, where each standard instance is paired with an unanswerable variant that has minimal semantic differences, in order for a reliable assessment. Evaluated upon 20 recent multi-modal LLMs, our results reveal that even the best-performing models like GPT-4o and Gemini Pro find it challenging to solve MuirBench, achieving 68.0% and 49.3% in accuracy. Open-source multimodal LLMs trained on single images can hardly generalize to multi-image questions, hovering below 33.3% in accuracy. These results highlight the importance of MuirBench in encouraging the community to develop multimodal LLMs that can look beyond a single image, suggesting potential pathways for future improvements.
Authors: Yiyuan Zhang, Handong Li, Jing Liu, Xiangyu Yue
Abstract: We propose to build omni-modal intelligence, which is capable of understanding any modality and learning universal representations. In specific, we propose a scalable pretraining paradigm, named Multimodal Context (MiCo), which can scale up the numbers of modalities and amount of data, together with the model parameters, in the pretraining process. With MiCo, the pretrained models show significant emergent abilities in multimodal learning, which are evaluated on the following tasks: i) single-modality perception benchmarks of 10 different modalities, ii) 25 cross-modality understanding tasks of retrieval, question-answering, captioning, and iii) 18 multimodal large language model benchmarks. Our models establish 37 new records for state-of-the-art performance. We hope that our research could contribute to the development of omni-modal intelligence. Code and Models are at https://github.com/invictus717/MiCo
Authors: Amil Dravid, Yossi Gandelsman, Kuan-Chieh Wang, Rameen Abdal, Gordon Wetzstein, Alexei A. Efros, Kfir Aberman
Abstract: We investigate the space of weights spanned by a large collection of customized diffusion models. We populate this space by creating a dataset of over 60,000 models, each of which is a base model fine-tuned to insert a different person's visual identity. We model the underlying manifold of these weights as a subspace, which we term weights2weights. We demonstrate three immediate applications of this space -- sampling, editing, and inversion. First, as each point in the space corresponds to an identity, sampling a set of weights from it results in a model encoding a novel identity. Next, we find linear directions in this space corresponding to semantic edits of the identity (e.g., adding a beard). These edits persist in appearance across generated samples. Finally, we show that inverting a single image into this space reconstructs a realistic identity, even if the input image is out of distribution (e.g., a painting). Our results indicate that the weight space of fine-tuned diffusion models behaves as an interpretable latent space of identities.
Authors: Lihe Yang, Bingyi Kang, Zilong Huang, Zhen Zhao, Xiaogang Xu, Jiashi Feng, Hengshuang Zhao
Abstract: This work presents Depth Anything V2. Without pursuing fancy techniques, we aim to reveal crucial findings to pave the way towards building a powerful monocular depth estimation model. Notably, compared with V1, this version produces much finer and more robust depth predictions through three key practices: 1) replacing all labeled real images with synthetic images, 2) scaling up the capacity of our teacher model, and 3) teaching student models via the bridge of large-scale pseudo-labeled real images. Compared with the latest models built on Stable Diffusion, our models are significantly more efficient (more than 10x faster) and more accurate. We offer models of different scales (ranging from 25M to 1.3B params) to support extensive scenarios. Benefiting from their strong generalization capability, we fine-tune them with metric depth labels to obtain our metric depth models. In addition to our models, considering the limited diversity and frequent noise in current test sets, we construct a versatile evaluation benchmark with precise annotations and diverse scenes to facilitate future research.
Authors: Duy-Kien Nguyen, Mahmoud Assran, Unnat Jain, Martin R. Oswald, Cees G. M. Snoek, Xinlei Chen
Abstract: This work does not introduce a new method. Instead, we present an interesting finding that questions the necessity of the inductive bias -- locality in modern computer vision architectures. Concretely, we find that vanilla Transformers can operate by directly treating each individual pixel as a token and achieve highly performant results. This is substantially different from the popular design in Vision Transformer, which maintains the inductive bias from ConvNets towards local neighborhoods (e.g. by treating each 16x16 patch as a token). We mainly showcase the effectiveness of pixels-as-tokens across three well-studied tasks in computer vision: supervised learning for object classification, self-supervised learning via masked autoencoding, and image generation with diffusion models. Although directly operating on individual pixels is less computationally practical, we believe the community must be aware of this surprising piece of knowledge when devising the next generation of neural architectures for computer vision.
Authors: Qihao Liu, Zhanpeng Zeng, Ju He, Qihang Yu, Xiaohui Shen, Liang-Chieh Chen
Abstract: This paper presents innovative enhancements to diffusion models by integrating a novel multi-resolution network and time-dependent layer normalization. Diffusion models have gained prominence for their effectiveness in high-fidelity image generation. While conventional approaches rely on convolutional U-Net architectures, recent Transformer-based designs have demonstrated superior performance and scalability. However, Transformer architectures, which tokenize input data (via "patchification"), face a trade-off between visual fidelity and computational complexity due to the quadratic nature of self-attention operations concerning token length. While larger patch sizes enable attention computation efficiency, they struggle to capture fine-grained visual details, leading to image distortions. To address this challenge, we propose augmenting the Diffusion model with the Multi-Resolution network (DiMR), a framework that refines features across multiple resolutions, progressively enhancing detail from low to high resolution. Additionally, we introduce Time-Dependent Layer Normalization (TD-LN), a parameter-efficient approach that incorporates time-dependent parameters into layer normalization to inject time information and achieve superior performance. Our method's efficacy is demonstrated on the class-conditional ImageNet generation benchmark, where DiMR-XL variants outperform prior diffusion models, setting new state-of-the-art FID scores of 1.70 on ImageNet 256 x 256 and 2.89 on ImageNet 512 x 512. Project page: https://qihao067.github.io/projects/DiMR
Authors: David McAllister, Songwei Ge, Jia-Bin Huang, David W. Jacobs, Alexei A. Efros, Aleksander Holynski, Angjoo Kanazawa
Abstract: Score distillation sampling (SDS) has proven to be an important tool, enabling the use of large-scale diffusion priors for tasks operating in data-poor domains. Unfortunately, SDS has a number of characteristic artifacts that limit its usefulness in general-purpose applications. In this paper, we make progress toward understanding the behavior of SDS and its variants by viewing them as solving an optimal-cost transport path from a source distribution to a target distribution. Under this new interpretation, these methods seek to transport corrupted images (source) to the natural image distribution (target). We argue that current methods' characteristic artifacts are caused by (1) linear approximation of the optimal path and (2) poor estimates of the source distribution. We show that calibrating the text conditioning of the source distribution can produce high-quality generation and translation results with little extra overhead. Our method can be easily applied across many domains, matching or beating the performance of specialized methods. We demonstrate its utility in text-to-2D, text-based NeRF optimization, translating paintings to real images, optical illusion generation, and 3D sketch-to-real. We compare our method to existing approaches for score distillation sampling and show that it can produce high-frequency details with realistic colors.
Authors: Muhammad Maaz, Hanoona Rasheed, Salman Khan, Fahad Khan
Abstract: Building on the advances of language models, Large Multimodal Models (LMMs) have contributed significant improvements in video understanding. While the current video LMMs utilize advanced Large Language Models (LLMs), they rely on either image or video encoders to process visual inputs, each of which has its own limitations. Image encoders excel at capturing rich spatial details from frame sequences but lack explicit temporal context, which can be important in videos with intricate action sequences. On the other hand, video encoders provide temporal context but are often limited by computational constraints that lead to processing only sparse frames at lower resolutions, resulting in reduced contextual and spatial understanding. To this end, we introduce VideoGPT+, which combines the complementary benefits of the image encoder (for detailed spatial understanding) and the video encoder (for global temporal context modeling). The model processes videos by dividing them into smaller segments and applies an adaptive pooling strategy on features extracted by both image and video encoders. Our architecture showcases improved performance across multiple video benchmarks, including VCGBench, MVBench and Zero-shot question-answering. Further, we develop 112K video-instruction set using a novel semi-automatic annotation pipeline which further improves the model performance. Additionally, to comprehensively evaluate video LMMs, we present VCGBench-Diverse, covering 18 broad video categories such as lifestyle, sports, science, gaming, and surveillance videos. This benchmark with 4,354 question-answer pairs evaluates the generalization of existing LMMs on dense video captioning, spatial and temporal understanding, and complex reasoning, ensuring comprehensive assessment across diverse video types and dynamics. Code: https://github.com/mbzuai-oryx/VideoGPT-plus.
Authors: Ankit Goyal, Valts Blukis, Jie Xu, Yijie Guo, Yu-Wei Chao, Dieter Fox
Abstract: In this work, we study how to build a robotic system that can solve multiple 3D manipulation tasks given language instructions. To be useful in industrial and household domains, such a system should be capable of learning new tasks with few demonstrations and solving them precisely. Prior works, like PerAct and RVT, have studied this problem, however, they often struggle with tasks requiring high precision. We study how to make them more effective, precise, and fast. Using a combination of architectural and system-level improvements, we propose RVT-2, a multitask 3D manipulation model that is 6X faster in training and 2X faster in inference than its predecessor RVT. RVT-2 achieves a new state-of-the-art on RLBench, improving the success rate from 65% to 82%. RVT-2 is also effective in the real world, where it can learn tasks requiring high precision, like picking up and inserting plugs, with just 10 demonstrations. Visual results, code, and trained model are provided at: https://robotic-view-transformer-2.github.io/.
Authors: Ayush Roy, Payel Pramanik, Sohom Ghosal, Daria Valenkova, Dmitrii Kaplun, Ram Sarkar
Abstract: Breast cancer is a major global health concern. Pathologists face challenges in analyzing complex features from pathological images, which is a time-consuming and labor-intensive task. Therefore, efficient computer-based diagnostic tools are needed for early detection and treatment planning. This paper presents a modified version of MultiResU-Net for histopathology image segmentation, which is selected as the backbone for its ability to analyze and segment complex features at multiple scales and ensure effective feature flow via skip connections. The modified version also utilizes the Gaussian distribution-based Attention Module (GdAM) to incorporate histopathology-relevant text information in a Gaussian distribution. The sampled features from the Gaussian text feature-guided distribution highlight specific spatial regions based on prior knowledge. Finally, using the Controlled Dense Residual Block (CDRB) on skip connections of MultiResU-Net, the information is transferred from the encoder layers to the decoder layers in a controlled manner using a scaling parameter derived from the extracted spatial features. We validate our approach on two diverse breast cancer histopathology image datasets: TNBC and MonuSeg, demonstrating superior segmentation performance compared to state-of-the-art methods. The code for our proposed model is available on https://github.com/AyushRoy2001/GRU-Net.
Authors: Zhongao Sun, Jiameng Li, Yuhan Wang, Jiarong Cheng, Qing Zhou, Chun Li
Abstract: Brain tumor segmentation remains a significant challenge, particularly in the context of multi-modal magnetic resonance imaging (MRI) where missing modality images are common in clinical settings, leading to reduced segmentation accuracy. To address this issue, we propose a novel strategy, which is called masked predicted pre-training, enabling robust feature learning from incomplete modality data. Additionally, in the fine-tuning phase, we utilize a knowledge distillation technique to align features between complete and missing modality data, simultaneously enhancing model robustness. Notably, we leverage the Holder pseudo-divergence instead of the KLD for distillation loss, offering improve mathematical interpretability and properties. Extensive experiments on the BRATS2018 and BRATS2020 datasets demonstrate significant performance enhancements compared to existing state-of-the-art methods.
Authors: Negar Kamali, Karyn Nakamura, Angelos Chatzimparmpas, Jessica Hullman, Matthew Groh
Abstract: The high level of photorealism in state-of-the-art diffusion models like Midjourney, Stable Diffusion, and Firefly makes it difficult for untrained humans to distinguish between real photographs and AI-generated images. To address this problem, we designed a guide to help readers develop a more critical eye toward identifying artifacts, inconsistencies, and implausibilities that often appear in AI-generated images. The guide is organized into five categories of artifacts and implausibilities: anatomical, stylistic, functional, violations of physics, and sociocultural. For this guide, we generated 138 images with diffusion models, curated 9 images from social media, and curated 42 real photographs. These images showcase the kinds of cues that prompt suspicion towards the possibility an image is AI-generated and why it is often difficult to draw conclusions about an image's provenance without any context beyond the pixels in an image. Human-perceptible artifacts are not always present in AI-generated images, but this guide reveals artifacts and implausibilities that often emerge. By drawing attention to these kinds of artifacts and implausibilities, we aim to better equip people to distinguish AI-generated images from real photographs in the future.
Authors: Kang-il Lee, Minbeom Kim, Seunghyun Yoon, Minsung Kim, Dongryeol Lee, Hyukhun Koh, Kyomin Jung
Abstract: Large Vision-Language Models (LVLMs) have demonstrated outstanding performance across various multimodal tasks. However, they suffer from a problem known as language prior, where responses are generated based solely on textual patterns while disregarding image information. Addressing the issue of language prior is crucial, as it can lead to undesirable biases or hallucinations when dealing with images that are out of training distribution. Despite its importance, current methods for accurately measuring language priors in LVLMs are poorly studied. Although existing benchmarks based on counterfactual or out-of-distribution images can partially be used to measure language priors, they fail to disentangle language priors from other confounding factors. To this end, we propose a new benchmark called VLind-Bench, which is the first benchmark specifically designed to measure the language priors, or blindness, of LVLMs. It not only includes tests on counterfactual images to assess language priors but also involves a series of tests to evaluate more basic capabilities such as commonsense knowledge, visual perception, and commonsense biases. For each instance in our benchmark, we ensure that all these basic tests are passed before evaluating the language priors, thereby minimizing the influence of other factors on the assessment. The evaluation and analysis of recent LVLMs in our benchmark reveal that almost all models exhibit a significant reliance on language priors, presenting a strong challenge in the field.
Authors: Matthieu Futeral, Armel Zebaze, Pedro Ortiz Suarez, Julien Abadji, R\'emi Lacroix, Cordelia Schmid, Rachel Bawden, Beno\^it Sagot
Abstract: Multimodal Large Language Models (mLLMs) are trained on a large amount of text-image data. While most mLLMs are trained on caption-like data only, Alayrac et al. [2022] showed that additionally training them on interleaved sequences of text and images can lead to the emergence of in-context learning capabilities. However, the dataset they used, M3W, is not public and is only in English. There have been attempts to reproduce their results but the released datasets are English-only. In contrast, current multilingual and multimodal datasets are either composed of caption-like only or medium-scale or fully private data. This limits mLLM research for the 7,000 other languages spoken in the world. We therefore introduce mOSCAR, to the best of our knowledge the first large-scale multilingual and multimodal document corpus crawled from the web. It covers 163 languages, 315M documents, 214B tokens and 1.2B images. We carefully conduct a set of filtering and evaluation steps to make sure mOSCAR is sufficiently safe, diverse and of good quality. We additionally train two types of multilingual model to prove the benefits of mOSCAR: (1) a model trained on a subset of mOSCAR and captioning data and (2) a model train on captioning data only. The model additionally trained on mOSCAR shows a strong boost in few-shot learning performance across various multilingual image-text tasks and benchmarks, confirming previous findings for English-only mLLMs.
Authors: Xinrui Yang, Zhuohan Wang, Anthony Hu
Abstract: Text-to-image models have shown remarkable progress in generating high-quality images from user-provided prompts. Despite this, the quality of these images varies due to the models' sensitivity to human language nuances. With advancements in large language models, there are new opportunities to enhance prompt design for image generation tasks. Existing research primarily focuses on optimizing prompts for direct interaction, while less attention is given to scenarios involving intermediary agents, like the Stable Diffusion model. This study proposes a Multi-Agent framework to optimize input prompts for text-to-image generation models. Central to this framework is a prompt generation mechanism that refines initial queries using dynamic instructions, which evolve through iterative performance feedback. High-quality prompts are then fed into a state-of-the-art text-to-image model. A professional prompts database serves as a benchmark to guide the instruction modifier towards generating high-caliber prompts. A scoring system evaluates the generated images, and an LLM generates new instructions based on calculated gradients. This iterative process is managed by the Upper Confidence Bound (UCB) algorithm and assessed using the Human Preference Score version 2 (HPS v2). Preliminary ablation studies highlight the effectiveness of various system components and suggest areas for future improvements.
Authors: Xinyun Liu, Chen Zhao
Abstract: Coronary artery disease (CAD) remains a prevalent cardiovascular condition, posing significant health risks worldwide. This pathology, characterized by plaque accumulation in coronary artery walls, leads to myocardial ischemia and various symptoms, including chest pain and shortness of breath. Accurate segmentation of coronary arteries from coronary computed tomography angiography (CCTA) images is crucial for diagnosis and treatment planning. Traditional segmentation methods face challenges in handling low-contrast images and complex anatomical structures. In this study, we propose an attention-guided, feature-aggregated 3D deep network (AGFA-Net) for coronary artery segmentation using CCTA images. AGFA-Net leverages attention mechanisms and feature refinement modules to capture salient features and enhance segmentation accuracy. Evaluation on a dataset comprising 1,000 CCTA scans demonstrates AGFA-Net's superior performance, achieving an average Dice coefficient similarity of 86.74% and a Hausdorff distance of 0.23 mm during 5-fold cross-validation. Ablation studies further validate the effectiveness of the proposed modules, highlighting their contributions to improved segmentation accuracy. Overall, AGFA-Net offers a robust and reliable solution for coronary artery segmentation, addressing challenges posed by varying vessel sizes, complex anatomies, and low image contrast.
Authors: Hao Liang, Chengjie, Kun Li, Xin Tian
Abstract: Hyperspectral image (HSI) denoising is an essential procedure for HSI applications. Unfortunately, the existing Transformer-based methods mainly focus on non-local modeling, neglecting the importance of locality in image denoising. Moreover, deep learning methods employ complex spectral learning mechanisms, thus introducing large computation costs. To address these problems, we propose a hybrid spatial-spectral denoising network (HSSD), in which we design a novel hybrid dual-path network inspired by CNN and Transformer characteristics, leading to capturing both local and non-local spatial details while suppressing noise efficiently. Furthermore, to reduce computational complexity, we adopt a simple but effective decoupling strategy that disentangles the learning of space and spectral channels, where multilayer perception with few parameters is utilized to learn the global correlations among spectra. The synthetic and real experiments demonstrate that our proposed method outperforms state-of-the-art methods on spatial and spectral reconstruction. The code and details are available on https://github.com/HLImg/HSSD.
Authors: Houze Liu, Iris Li, Yaxin Liang, Dan Sun, Yining Yang, Haowei Yang
Abstract: Neural networks with relatively shallow layers and simple structures may have limited ability in accurately identifying pneumonia. In addition, deep neural networks also have a large demand for computing resources, which may cause convolutional neural networks to be unable to be implemented on terminals. Therefore, this paper will carry out the optimal classification of convolutional neural networks. Firstly, according to the characteristics of pneumonia images, AlexNet and InceptionV3 were selected to obtain better image recognition results. Combining the features of medical images, the forward neural network with deeper and more complex structure is learned. Finally, knowledge extraction technology is used to extract the obtained data into the AlexNet model to achieve the purpose of improving computing efficiency and reducing computing costs. The results showed that the prediction accuracy, specificity, and sensitivity of the trained AlexNet model increased by 4.25 percentage points, 7.85 percentage points, and 2.32 percentage points, respectively. The graphics processing usage has decreased by 51% compared to the InceptionV3 mode.
Authors: Tairan He, Zhengyi Luo, Xialin He, Wenli Xiao, Chong Zhang, Weinan Zhang, Kris Kitani, Changliu Liu, Guanya Shi
Abstract: We present OmniH2O (Omni Human-to-Humanoid), a learning-based system for whole-body humanoid teleoperation and autonomy. Using kinematic pose as a universal control interface, OmniH2O enables various ways for a human to control a full-sized humanoid with dexterous hands, including using real-time teleoperation through VR headset, verbal instruction, and RGB camera. OmniH2O also enables full autonomy by learning from teleoperated demonstrations or integrating with frontier models such as GPT-4. OmniH2O demonstrates versatility and dexterity in various real-world whole-body tasks through teleoperation or autonomy, such as playing multiple sports, moving and manipulating objects, and interacting with humans. We develop an RL-based sim-to-real pipeline, which involves large-scale retargeting and augmentation of human motion datasets, learning a real-world deployable policy with sparse sensor input by imitating a privileged teacher policy, and reward designs to enhance robustness and stability. We release the first humanoid whole-body control dataset, OmniH2O-6, containing six everyday tasks, and demonstrate humanoid whole-body skill learning from teleoperated datasets.
Authors: Yuhan Quan, Huan ZHao, Jinfeng Yi, Yuqiang Chen
Abstract: CAD (Computer-Aided Design) plays a crucial role in mechanical industry, where large numbers of similar-shaped CAD parts are often created. Efficiently reusing these parts is key to reducing design and production costs for enterprises. Retrieval systems are vital for achieving CAD reuse, but the complex shapes of CAD models are difficult to accurately describe using text or keywords, making traditional retrieval methods ineffective. While existing representation learning approaches have been developed for CAD, manually labeling similar samples in these methods is expensive. Additionally, CAD models' unique parameterized data structure presents challenges for applying existing 3D shape representation learning techniques directly. In this work, we propose GC-CAD, a self-supervised contrastive graph neural network-based method for mechanical CAD retrieval that directly models parameterized CAD raw files. GC-CAD consists of two key modules: structure-aware representation learning and contrastive graph learning framework. The method leverages graph neural networks to extract both geometric and topological information from CAD models, generating feature representations. We then introduce a simple yet effective contrastive graph learning framework approach, enabling the model to train without manual labels and generate retrieval-ready representations. Experimental results on four datasets including human evaluation demonstrate that the proposed method achieves significant accuracy improvements and up to 100 times efficiency improvement over the baseline methods.
Authors: Jingyuan Xia, Zhixiong Yang, Shengxi Li, Shuanghui Zhang, Yaowen Fu, Deniz G\"und\"uz, Xiang Li
Abstract: Learning-based approaches have witnessed great successes in blind single image super-resolution (SISR) tasks, however, handcrafted kernel priors and learning based kernel priors are typically required. In this paper, we propose a Meta-learning and Markov Chain Monte Carlo (MCMC) based SISR approach to learn kernel priors from organized randomness. In concrete, a lightweight network is adopted as kernel generator, and is optimized via learning from the MCMC simulation on random Gaussian distributions. This procedure provides an approximation for the rational blur kernel, and introduces a network-level Langevin dynamics into SISR optimization processes, which contributes to preventing bad local optimal solutions for kernel estimation. Meanwhile, a meta-learning-based alternating optimization procedure is proposed to optimize the kernel generator and image restorer, respectively. In contrast to the conventional alternating minimization strategy, a meta-learning-based framework is applied to learn an adaptive optimization strategy, which is less-greedy and results in better convergence performance. These two procedures are iteratively processed in a plug-and-play fashion, for the first time, realizing a learning-based but plug-and-play blind SISR solution in unsupervised inference. Extensive simulations demonstrate the superior performance and generalization ability of the proposed approach when comparing with state-of-the-arts on synthesis and real-world datasets. The code is available at https://github.com/XYLGroup/MLMC.
Authors: Krzysztof Wolski, Adarsh Djeacoumar, Alireza Javanmardi, Hans-Peter Seidel, Christian Theobalt, Guillaume Cordonnier, Karol Myszkowski, George Drettakis, Xingang Pan, Thomas Leimk\"uhler
Abstract: The real world exhibits rich structure and detail across many scales of observation. It is difficult, however, to capture and represent a broad spectrum of scales using ordinary images. We devise a novel paradigm for learning a representation that captures an orders-of-magnitude variety of scales from an unstructured collection of ordinary images. We treat this collection as a distribution of scale-space slices to be learned using adversarial training, and additionally enforce coherency across slices. Our approach relies on a multiscale generator with carefully injected procedural frequency content, which allows to interactively explore the emerging continuous scale space. Training across vastly different scales poses challenges regarding stability, which we tackle using a supervision scheme that involves careful sampling of scales. We show that our generator can be used as a multiscale generative model, and for reconstructions of scale spaces from unstructured patches. Significantly outperforming the state of the art, we demonstrate zoom-in factors of up to 256x at high quality and scale consistency.
Authors: Preetum Nakkiran, Arwen Bradley, Hattie Zhou, Madhu Advani
Abstract: We present an accessible first course on diffusion models and flow matching for machine learning, aimed at a technical audience with no diffusion experience. We try to simplify the mathematical details as much as possible (sometimes heuristically), while retaining enough precision to derive correct algorithms.
Authors: Zahraa Al Sahili, Ioannis Patras, Matthew Purver
Abstract: In the domain of text-to-image generative models, the inadvertent propagation of biases inherent in training datasets poses significant ethical challenges, particularly in the generation of socially sensitive content. This paper introduces EquiPrompt, a novel method employing Chain of Thought (CoT) reasoning to reduce biases in text-to-image generative models. EquiPrompt uses iterative bootstrapping and bias-aware exemplar selection to balance creativity and ethical responsibility. It integrates iterative reasoning refinement with controlled evaluation techniques, addressing zero-shot CoT issues in sensitive contexts. Experiments on several generation tasks show EquiPrompt effectively lowers bias while maintaining generative quality, advancing ethical AI and socially responsible creative processes.Code will be publically available.
Authors: Melvin Wong, Thiago Rios, Stefan Menzel, Yew Soon Ong
Abstract: Engineering design optimization requires an efficient combination of a 3D shape representation, an optimization algorithm, and a design performance evaluation method, which is often computationally expensive. We present a prompt evolution design optimization (PEDO) framework contextualized in a vehicle design scenario that leverages a vision-language model for penalizing impractical car designs synthesized by a generative model. The backbone of our framework is an evolutionary strategy coupled with an optimization objective function that comprises a physics-based solver and a vision-language model for practical or functional guidance in the generated car designs. In the prompt evolutionary search, the optimizer iteratively generates a population of text prompts, which embed user specifications on the aerodynamic performance and visual preferences of the 3D car designs. Then, in addition to the computational fluid dynamics simulations, the pre-trained vision-language model is used to penalize impractical designs and, thus, foster the evolutionary algorithm to seek more viable designs. Our investigations on a car design optimization problem show a wide spread of potential car designs generated at the early phase of the search, which indicates a good diversity of designs in the initial populations, and an increase of over 20\% in the probability of generating practical designs compared to a baseline framework without using a vision-language model. Visual inspection of the designs against the performance results demonstrates prompt evolution as a very promising paradigm for finding novel designs with good optimization performance while providing ease of use in specifying design specifications and preferences via a natural language interface.
Authors: A B M Ashikur Rahman, Saeed Anwar, Muhammad Usman, Ajmal Mian
Abstract: Large Language Models (LLMs) have demonstrated remarkable capabilities, revolutionizing the integration of AI in daily life applications. However, they are prone to hallucinations, generating claims that contradict established facts, deviating from prompts, and producing inconsistent responses when the same prompt is presented multiple times. Addressing these issues is challenging due to the lack of comprehensive and easily assessable benchmark datasets. Most existing datasets are small and rely on multiple-choice questions, which are inadequate for evaluating the generative prowess of LLMs. To measure hallucination in LLMs, this paper introduces a comprehensive benchmark dataset comprising over 75,000 prompts across eight domains. These prompts are designed to elicit definitive, concise, and informative answers. The dataset is divided into two segments: one publicly available for testing and assessing LLM performance and a hidden segment for benchmarking various LLMs. In our experiments, we tested six LLMs-GPT-3.5, LLama 2, LLama 3, Gemini, Mixtral, and Zephyr-revealing that overall factual hallucination ranges from 59% to 82% on the public dataset and 57% to 76% in the hidden benchmark. Prompt misalignment hallucination ranges from 6% to 95% in the public dataset and 17% to 94% in the hidden counterpart. Average consistency ranges from 21% to 61% and 22% to 63%, respectively. Domain-wise analysis shows that LLM performance significantly deteriorates when asked for specific numeric information while performing moderately with person, location, and date queries. Our dataset demonstrates its efficacy and serves as a comprehensive benchmark for LLM performance evaluation. Our dataset and LLMs responses are available at \href{https://github.com/ashikiut/DefAn}{https://github.com/ashikiut/DefAn}.
URLs: https://github.com/ashikiut/DefAn, https://github.com/ashikiut/DefAn
Authors: Orchid Chetia Phukan, Priyabrata Mallick, Swarup Ranjan Behera, Aalekhya Satya Narayani, Arun Balaji Buduru, Rajesh Sharma
Abstract: In this paper, we work towards extending Audio-Visual Question Answering (AVQA) to multilingual settings. Existing AVQA research has predominantly revolved around English and replicating it for addressing AVQA in other languages requires a substantial allocation of resources. As a scalable solution, we leverage machine translation and present two multilingual AVQA datasets for eight languages created from existing benchmark AVQA datasets. This prevents extra human annotation efforts of collecting questions and answers manually. To this end, we propose, MERA framework, by leveraging state-of-the-art (SOTA) video, audio, and textual foundation models for AVQA in multiple languages. We introduce a suite of models namely MERA-L, MERA-C, MERA-T with varied model architectures to benchmark the proposed datasets. We believe our work will open new research directions and act as a reference benchmark for future works in multilingual AVQA.
Authors: Ludvig Ericson, Patric Jensfelt
Abstract: In this paper, we tackle the challenge of predicting the unseen walls of a partially observed environment as a set of 2D line segments, conditioned on occupancy grids integrated along the trajectory of a 360{\deg} LIDAR sensor. A dataset of such occupancy grids and their corresponding target wall segments is collected by navigating a virtual robot between a set of randomly sampled waypoints in a collection of office-scale floor plans from a university campus. The line segment prediction task is formulated as an autoregressive sequence prediction task, and an attention-based deep network is trained on the dataset. The sequence-based autoregressive formulation is evaluated through predicted information gain, as in frontier-based autonomous exploration, demonstrating significant improvements over both non-predictive estimation and convolution-based image prediction found in the literature. Ablations on key components are evaluated, as well as sensor range and the occupancy grid's metric area. Finally, model generality is validated by predicting walls in a novel floor plan reconstructed on-the-fly in a real-world office environment.
Authors: Soufiane Belharbi, Mara KM Whitford, Phuong Hoang, Shakeeb Murtaza, Luke McCaffrey, Eric Granger
Abstract: Confocal fluorescence microscopy is one of the most accessible and widely used imaging techniques for the study of biological processes. Scanning confocal microscopy allows the capture of high-quality images from 3D samples, yet suffers from well-known limitations such as photobleaching and phototoxicity of specimens caused by intense light exposure, which limits its use in some applications, especially for living cells. Cellular damage can be alleviated by changing imaging parameters to reduce light exposure, often at the expense of image quality. Machine/deep learning methods for single-image super-resolution (SISR) can be applied to restore image quality by upscaling lower-resolution (LR) images to produce high-resolution images (HR). These SISR methods have been successfully applied to photo-realistic images due partly to the abundance of publicly available data. In contrast, the lack of publicly available data partly limits their application and success in scanning confocal microscopy. In this paper, we introduce a large scanning confocal microscopy dataset named SR-CACO-2 that is comprised of low- and high-resolution image pairs marked for three different fluorescent markers. It allows the evaluation of performance of SISR methods on three different upscaling levels (X2, X4, X8). SR-CACO-2 contains the human epithelial cell line Caco-2 (ATCC HTB-37), and it is composed of 22 tiles that have been translated in the form of 9,937 image patches for experiments with SISR methods. Given the new SR-CACO-2 dataset, we also provide benchmarking results for 15 state-of-the-art methods that are representative of the main SISR families. Results show that these methods have limited success in producing high-resolution textures, indicating that SR-CACO-2 represents a challenging problem. Our dataset, code and pretrained weights are available: https://github.com/sbelharbi/sr-caco-2.
Authors: Yuchi Liu, Yifan Sun, Jingdong Wang, Liang Zheng
Abstract: This paper evaluates the generalization ability of classification models on out-of-distribution test sets without depending on ground truth labels. Common approaches often calculate an unsupervised metric related to a specific model property, like confidence or invariance, which correlates with out-of-distribution accuracy. However, these metrics are typically computed for each test sample individually, leading to potential issues caused by spurious model responses, such as overly high or low confidence. To tackle this challenge, we propose incorporating responses from neighboring test samples into the correctness assessment of each individual sample. In essence, if a model consistently demonstrates high correctness scores for nearby samples, it increases the likelihood of correctly predicting the target sample, and vice versa. The resulting scores are then averaged across all test samples to provide a holistic indication of model accuracy. Developed under the vicinal risk formulation, this approach, named vicinal risk proxy (VRP), computes accuracy without relying on labels. We show that applying the VRP method to existing generalization indicators, such as average confidence and effective invariance, consistently improves over these baselines both methodologically and experimentally. This yields a stronger correlation with model accuracy, especially on challenging out-of-distribution test sets.
Authors: Th\'eo Moutakanni, Maxime Oquab, Marc Szafraniec, Maria Vakalopoulou, Piotr Bojanowski
Abstract: Self-Supervised learning (SSL) with Joint-Embedding Architectures (JEA) has led to outstanding performances. All instantiations of this paradigm were trained using strong and well-established hand-crafted data augmentations, leading to the general belief that they are required for the proper training and performance of such models. On the other hand, generative reconstruction-based models such as BEIT and MAE or Joint-Embedding Predictive Architectures such as I-JEPA have shown strong performance without using data augmentations except masking. In this work, we challenge the importance of invariance and data-augmentation in JEAs at scale. By running a case-study on a recent SSL foundation model - DINOv2 - we show that strong image representations can be obtained with JEAs and only cropping without resizing provided the training data is large enough, reaching state-of-the-art results and using the least amount of augmentation in the literature. Through this study, we also discuss the impact of compute constraints on the outcomes of experimental deep learning research, showing that they can lead to very different conclusions.
Authors: Yuhang Wu, Wenmeng Yu, Yean Cheng, Yan Wang, Xiaohan Zhang, Jiazheng Xu, Ming Ding, Yuxiao Dong
Abstract: Evaluating the alignment capabilities of large Vision-Language Models (VLMs) is essential for determining their effectiveness as helpful assistants. However, existing benchmarks primarily focus on basic abilities using nonverbal methods, such as yes-no and multiple-choice questions. In this paper, we address this gap by introducing AlignMMBench, a comprehensive alignment benchmark specifically designed for emerging Chinese VLMs. This benchmark is meticulously curated from real-world scenarios and Chinese Internet sources, encompassing thirteen specific tasks across three categories, and includes both single-turn and multi-turn dialogue scenarios. Incorporating a prompt rewrite strategy, AlignMMBench encompasses 1,054 images and 4,978 question-answer pairs. To facilitate the evaluation pipeline, we propose CritiqueVLM, a rule-calibrated evaluator that exceeds GPT-4's evaluation ability. Finally, we report the performance of representative VLMs on AlignMMBench, offering insights into the capabilities and limitations of different VLM architectures. All evaluation codes and data are available on https://alignmmbench.github.io.
Authors: Zhuolin Fu
Abstract: In this paper, we show how Transformers can be interpreted as dense Expectation-Maximization algorithms performed on Bayesian Nets. Based on the above interpretation, we propose a new model design paradigm, namely Vertical LoRA (VLoRA), which reduces the parameter count dramatically while preserving performance. In VLoRA, a model consists of layers, each of which recursively learns an increment based on the previous layer. We then apply LoRA decomposition to the increments. VLoRA works on the base model, which is orthogonal to LoRA, meaning they can be used together. We do experiments on various tasks and models. The results show that 1) with VLoRA, the Transformer model parameter count can be reduced dramatically and 2) the performance of the original model is preserved. The source code is available at \url{https://github.com/neverUseThisName/vlora}
Authors: Meng Wang, Tian Lin, Kai Yu, Aidi Lin, Yuanyuan Peng, Lianyu Wang, Cheng Chen, Ke Zou, Huiyu Liang, Man Chen, Xue Yao, Meiqin Zhang, Binwei Huang, Chaoxin Zheng, Wei Chen, Yilong Luo, Yifan Chen, Jingcheng Wang, Yih Chung Tham, Dianbo Liu, Wendy Wong, Sahil Thakur, Beau Fenner, Yanda Meng, Yukun Zhou, Zehua Jiang, Minghui Qiu, Changqing Zhang, Xinjian Chen, Sophia Y. Wang, Cecilia S. Lee, Lucia Sobrin, Pearse A. Keane, Ching-Yu Cheng, Haoyu Chen, Huazhu Fu
Abstract: The current retinal artificial intelligence models were trained using data with a limited category of diseases and limited knowledge. In this paper, we present a retinal vision-language foundation model (RetiZero) with knowledge of over 400 fundus diseases. Specifically, we collected 341,896 fundus images paired with text descriptions from 29 publicly available datasets, 180 ophthalmic books, and online resources, encompassing over 400 fundus diseases across multiple countries and ethnicities. RetiZero achieved outstanding performance across various downstream tasks, including zero-shot retinal disease recognition, image-to-image retrieval, internal domain and cross-domain retinal disease classification, and few-shot fine-tuning. Specially, in the zero-shot scenario, RetiZero achieved a Top5 score of 0.8430 and 0.7561 on 15 and 52 fundus diseases respectively. In the image-retrieval task, RetiZero achieved a Top5 score of 0.9500 and 0.8860 on 15 and 52 retinal diseases respectively. Furthermore, clinical evaluations by ophthalmology experts from different countries demonstrate that RetiZero can achieve performance comparable to experienced ophthalmologists using zero-shot and image retrieval methods without requiring model retraining. These capabilities of retinal disease identification strengthen our RetiZero foundation model in clinical implementation.
Authors: Qijun Gan, Song Wang, Shengtao Wu, Jianke Zhu
Abstract: Recently, artificial intelligence techniques for education have been received increasing attentions, while it still remains an open problem to design the effective music instrument instructing systems. Although key presses can be directly derived from sheet music, the transitional movements among key presses require more extensive guidance in piano performance. In this work, we construct a piano-hand motion generation benchmark to guide hand movements and fingerings for piano playing. To this end, we collect an annotated dataset, PianoMotion10M, consisting of 116 hours of piano playing videos from a bird's-eye view with 10 million annotated hand poses. We also introduce a powerful baseline model that generates hand motions from piano audios through a position predictor and a position-guided gesture generator. Furthermore, a series of evaluation metrics are designed to assess the performance of the baseline model, including motion similarity, smoothness, positional accuracy of left and right hands, and overall fidelity of movement distribution. Despite that piano key presses with respect to music scores or audios are already accessible, PianoMotion10M aims to provide guidance on piano fingering for instruction purposes. The dataset and source code can be accessed at https://agnjason.github.io/PianoMotion-page.
Authors: Stefan P. Hein, Manuel Schultheiss, Andrei Gafita, Raphael Zaum, Farid Yagubbayli, Isabel Rauscher, Matthias Eiber, Franz Pfeiffer, Wolfgang A. Weber
Abstract: Assessing tumor response to systemic therapies is one of the main applications of PET/CT. Routinely, only a small subset of index lesions out of multiple lesions is analyzed. However, this operator dependent selection may bias the results due to possible significant inter-metastatic heterogeneity of response to therapy. Automated, AI based approaches for lesion tracking hold promise in enabling the analysis of many more lesions and thus providing a better assessment of tumor response. This work introduces a Siamese CNN approach for lesion tracking between PET/CT scans. Our approach is applied on the laborious task of tracking a high number of bone lesions in full-body baseline and follow-up [68Ga]Ga- or [18F]F-PSMA PET/CT scans after two cycles of [177Lu]Lu-PSMA therapy of metastatic castration resistant prostate cancer patients. Data preparation includes lesion segmentation and affine registration. Our algorithm extracts suitable lesion patches and forwards them into a Siamese CNN trained to classify the lesion patch pairs as corresponding or non-corresponding lesions. Experiments have been performed with different input patch types and a Siamese network in 2D and 3D. The CNN model successfully learned to classify lesion assignments, reaching a lesion tracking accuracy of 83 % in its best configuration with an AUC = 0.91. For remaining lesions the pipeline accomplished a re-identification rate of 89 %. We proved that a CNN may facilitate the tracking of multiple lesions in PSMA PET/CT scans. Future clinical studies are necessary if this improves the prediction of the outcome of therapies.
Authors: Federico Spagnolo, Nataliia Molchanova, Roger Schaer, Meritxell Bach Cuadra, Mario Ocampo Pineda, Lester Melie-Garcia, Cristina Granziera, Vincent Andrearczyk, Adrien Depeursinge
Abstract: In recent years, explainable methods for artificial intelligence (XAI) have tried to reveal and describe models' decision mechanisms in the case of classification tasks. However, XAI for semantic segmentation and in particular for single instances has been little studied to date. Understanding the process underlying automatic segmentation of single instances is crucial to reveal what information was used to detect and segment a given object of interest. In this study, we proposed two instance-level explanation maps for semantic segmentation based on SmoothGrad and Grad-CAM++ methods. Then, we investigated their relevance for the detection and segmentation of white matter lesions (WML), a magnetic resonance imaging (MRI) biomarker in multiple sclerosis (MS). 687 patients diagnosed with MS for a total of 4043 FLAIR and MPRAGE MRI scans were collected at the University Hospital of Basel, Switzerland. Data were randomly split into training, validation and test sets to train a 3D U-Net for MS lesion segmentation. We observed 3050 true positive (TP), 1818 false positive (FP), and 789 false negative (FN) cases. We generated instance-level explanation maps for semantic segmentation, by developing two XAI methods based on SmoothGrad and Grad-CAM++. We investigated: 1) the distribution of gradients in saliency maps with respect to both input MRI sequences; 2) the model's response in the case of synthetic lesions; 3) the amount of perilesional tissue needed by the model to segment a lesion. Saliency maps (based on SmoothGrad) in FLAIR showed positive values inside a lesion and negative in its neighborhood. Peak values of saliency maps generated for these four groups of volumes presented distributions that differ significantly from one another, suggesting a quantitative nature of the proposed saliency. Contextual information of 7mm around the lesion border was required for their segmentation.
Authors: Hoang Phan, Lam Tran, Quyen Tran, Trung Le
Abstract: Prior Unsupervised Domain Adaptation (UDA) methods often aim to train a domain-invariant feature extractor, which may hinder the model from learning sufficiently discriminative features. To tackle this, a line of works based on prompt learning leverages the power of large-scale pre-trained vision-language models to learn both domain-invariant and specific features through a set of domain-agnostic and domain-specific learnable prompts. Those studies typically enforce invariant constraints on representation, output, or prompt space to learn such prompts. Differently, we cast UDA as a multiple-objective optimization problem in which each objective is represented by a domain loss. Under this new framework, we propose aligning per-objective gradients to foster consensus between them. Additionally, to prevent potential overfitting when fine-tuning this deep learning architecture, we penalize the norm of these gradients. To achieve these goals, we devise a practical gradient update procedure that can work under both single-source and multi-source UDA. Empirically, our method consistently surpasses other prompt-based baselines by a large margin on different UDA benchmarks
Authors: Rylan Schaeffer, Victor Lecomte, Dhruv Bhandarkar Pai, Andres Carranza, Berivan Isik, Alyssa Unell, Mikail Khona, Thomas Yerxa, Yann LeCun, SueYeon Chung, Andrey Gromov, Ravid Shwartz-Ziv, Sanmi Koyejo
Abstract: Maximum Manifold Capacity Representations (MMCR) is a recent multi-view self-supervised learning (MVSSL) method that matches or surpasses other leading MVSSL methods. MMCR is intriguing because it does not fit neatly into any of the commonplace MVSSL lineages, instead originating from a statistical mechanical perspective on the linear separability of data manifolds. In this paper, we seek to improve our understanding and our utilization of MMCR. To better understand MMCR, we leverage tools from high dimensional probability to demonstrate that MMCR incentivizes alignment and uniformity of learned embeddings. We then leverage tools from information theory to show that such embeddings maximize a well-known lower bound on mutual information between views, thereby connecting the geometric perspective of MMCR to the information-theoretic perspective commonly discussed in MVSSL. To better utilize MMCR, we mathematically predict and experimentally confirm non-monotonic changes in the pretraining loss akin to double descent but with respect to atypical hyperparameters. We also discover compute scaling laws that enable predicting the pretraining loss as a function of gradients steps, batch size, embedding dimension and number of views. We then show that MMCR, originally applied to image data, is performant on multimodal image-text data. By more deeply understanding the theoretical and empirical behavior of MMCR, our work reveals insights on improving MVSSL methods.
Authors: Lukas Thede, Karsten Roth, Olivier J. H\'enaff, Matthias Bethge, Zeynep Akata
Abstract: With the advent and recent ubiquity of foundation models, continual learning (CL) has recently shifted from continual training from scratch to the continual adaptation of pretrained models, seeing particular success on rehearsal-free CL benchmarks (RFCL). To achieve this, most proposed methods adapt and restructure parameter-efficient finetuning techniques (PEFT) to suit the continual nature of the problem. Based most often on input-conditional query-mechanisms or regularizations on top of prompt- or adapter-based PEFT, these PEFT-style RFCL (P-RFCL) approaches report peak performances; often convincingly outperforming existing CL techniques. However, on the other end, critical studies have recently highlighted competitive results by training on just the first task or via simple non-parametric baselines. Consequently, questions arise about the relationship between methodological choices in P-RFCL and their reported high benchmark scores. In this work, we tackle these questions to better understand the true drivers behind strong P-RFCL performances, their placement w.r.t. recent first-task adaptation studies, and their relation to preceding CL standards such as EWC or SI. In particular, we show: (1) P-RFCL techniques relying on input-conditional query mechanisms work not because, but rather despite them by collapsing towards standard PEFT shortcut solutions. (2) Indeed, we show how most often, P-RFCL techniques can be matched by a simple and lightweight PEFT baseline. (3) Using this baseline, we identify the implicit bound on tunable parameters when deriving RFCL approaches from PEFT methods as a potential denominator behind P-RFCL efficacy. Finally, we (4) better disentangle continual versus first-task adaptation, and (5) motivate standard RFCL techniques s.a. EWC or SI in light of recent P-RFCL methods.
Authors: Baiang Li, Sizhuo Ma, Yanhong Zeng, Xiaogang Xu, Youqing Fang, Zhao Zhang, Jian Wang, Kai Chen
Abstract: Capturing High Dynamic Range (HDR) scenery using 8-bit cameras often suffers from over-/underexposure, loss of fine details due to low bit-depth compression, skewed color distributions, and strong noise in dark areas. Traditional LDR image enhancement methods primarily focus on color mapping, which enhances the visual representation by expanding the image's color range and adjusting the brightness. However, these approaches fail to effectively restore content in dynamic range extremes, which are regions with pixel values close to 0 or 255. To address the full scope of challenges in HDR imaging and surpass the limitations of current models, we propose a novel two-stage approach. The first stage maps the color and brightness to an appropriate range while keeping the existing details, and the second stage utilizes a diffusion prior to generate content in dynamic range extremes lost during capture. This generative refinement module can also be used as a plug-and-play module to enhance and complement existing LDR enhancement models. The proposed method markedly improves the quality and details of LDR images, demonstrating superior performance through rigorous experimental validation. The project page is at https://sagiri0208.github.io
Authors: Jong-Ik Park, Sihoon Seong, JunKyu Lee, Cheol-Ho Hong
Abstract: Tabular data from IIoT devices are typically analyzed using decision tree-based machine learning techniques, which struggle with high-dimensional and numeric data. To overcome these limitations, techniques converting tabular data into images have been developed, leveraging the strengths of image-based deep learning approaches such as Convolutional Neural Networks. These methods cluster similar features into distinct image areas with fixed sizes, regardless of the number of features, resembling actual photographs. However, this increases the possibility of overfitting, as similar features, when selected carefully in a tabular format, are often discarded to prevent this issue. Additionally, fixed image sizes can lead to wasted pixels with fewer features, resulting in computational inefficiency. We introduce Vortex Feature Positioning (VFP) to address these issues. VFP arranges features based on their correlation, spacing similar ones in a vortex pattern from the image center, with the image size determined by the attribute count. VFP outperforms traditional machine learning methods and existing conversion techniques in tests across seven datasets with varying real-valued attributes.
Authors: Mingyang Song, Yang Zhang, Tun\c{c} O. Ayd{\i}n, Elham Amin Mansour, Christopher Schroers
Abstract: Noise synthesis is a challenging low-level vision task aiming to generate realistic noise given a clean image along with the camera settings. To this end, we propose an effective generative model which utilizes clean features as guidance followed by noise injections into the network. Specifically, our generator follows a UNet-like structure with skip connections but without downsampling and upsampling layers. Firstly, we extract deep features from a clean image as the guidance and concatenate a Gaussian noise map to the transition point between the encoder and decoder as the noise source. Secondly, we propose noise synthesis blocks in the decoder in each of which we inject Gaussian noise to model the noise characteristics. Thirdly, we propose to utilize an additional Style Loss and demonstrate that this allows better noise characteristics supervision in the generator. Through a number of new experiments, we evaluate the temporal variance and the spatial correlation of the generated noise which we hope can provide meaningful insights for future works. Finally, we show that our proposed approach outperforms existing methods for synthesizing camera noise.
Authors: Kim Hoang Tran, Anh Duy Le Dinh, Tien Phat Nguyen, Thinh Phan, Pha Nguyen, Khoa Luu, Donald Adjeroh, Gianfranco Doretto, Ngan Hoang Le
Abstract: Despite recent significant progress, Multi-Object Tracking (MOT) faces limitations such as reliance on prior knowledge and predefined categories and struggles with unseen objects. To address these issues, Generic Multiple Object Tracking (GMOT) has emerged as an alternative approach, requiring less prior information. However, current GMOT methods often rely on initial bounding boxes and struggle to handle variations in factors such as viewpoint, lighting, occlusion, and scale, among others. Our contributions commence with the introduction of the \textit{Referring GMOT dataset} a collection of videos, each accompanied by detailed textual descriptions of their attributes. Subsequently, we propose $\mathtt{Z-GMOT}$, a cutting-edge tracking solution capable of tracking objects from \textit{never-seen categories} without the need of initial bounding boxes or predefined categories. Within our $\mathtt{Z-GMOT}$ framework, we introduce two novel components: (i) $\mathtt{iGLIP}$, an improved Grounded language-image pretraining, for accurately detecting unseen objects with specific characteristics. (ii) $\mathtt{MA-SORT}$, a novel object association approach that adeptly integrates motion and appearance-based matching strategies to tackle the complex task of tracking objects with high similarity. Our contributions are benchmarked through extensive experiments conducted on the Referring GMOT dataset for GMOT task. Additionally, to assess the generalizability of the proposed $\mathtt{Z-GMOT}$, we conduct ablation studies on the DanceTrack and MOT20 datasets for the MOT task. Our dataset, code, and models are released at: https://fsoft-aic.github.io/Z-GMOT.
Authors: Hu Wang, Yuanhong Chen, Congbo Ma, Jodie Avery, Louise Hull, Gustavo Carneiro
Abstract: The missing modality issue is critical but non-trivial to be solved by multi-modal models. Current methods aiming to handle the missing modality problem in multi-modal tasks, either deal with missing modalities only during evaluation or train separate models to handle specific missing modality settings. In addition, these models are designed for specific tasks, so for example, classification models are not easily adapted to segmentation tasks and vice versa. In this paper, we propose the Shared-Specific Feature Modelling (ShaSpec) method that is considerably simpler and more effective than competing approaches that address the issues above. ShaSpec is designed to take advantage of all available input modalities during training and evaluation by learning shared and specific features to better represent the input data. This is achieved from a strategy that relies on auxiliary tasks based on distribution alignment and domain classification, in addition to a residual feature fusion procedure. Also, the design simplicity of ShaSpec enables its easy adaptation to multiple tasks, such as classification and segmentation. Experiments are conducted on both medical image segmentation and computer vision classification, with results indicating that ShaSpec outperforms competing methods by a large margin. For instance, on BraTS2018, ShaSpec improves the SOTA by more than 3% for enhancing tumour, 5% for tumour core and 3% for whole tumour. The code repository address is https://github.com/billhhh/ShaSpec/.
Authors: Sudarshan Babu, Richard Liu, Avery Zhou, Michael Maire, Greg Shakhnarovich, Rana Hanocka
Abstract: We introduce HyperFields, a method for generating text-conditioned Neural Radiance Fields (NeRFs) with a single forward pass and (optionally) some fine-tuning. Key to our approach are: (i) a dynamic hypernetwork, which learns a smooth mapping from text token embeddings to the space of NeRFs; (ii) NeRF distillation training, which distills scenes encoded in individual NeRFs into one dynamic hypernetwork. These techniques enable a single network to fit over a hundred unique scenes. We further demonstrate that HyperFields learns a more general map between text and NeRFs, and consequently is capable of predicting novel in-distribution and out-of-distribution scenes -- either zero-shot or with a few finetuning steps. Finetuning HyperFields benefits from accelerated convergence thanks to the learned general map, and is capable of synthesizing novel scenes 5 to 10 times faster than existing neural optimization-based methods. Our ablation experiments show that both the dynamic architecture and NeRF distillation are critical to the expressivity of HyperFields.
Authors: Haosen Yang, Chuofan Ma, Bin Wen, Yi Jiang, Zehuan Yuan, Xiatian Zhu
Abstract: Understanding the semantics of individual regions or patches of unconstrained images, such as open-world object detection, remains a critical yet challenging task in computer vision. Building on the success of powerful image-level vision-language (ViL) foundation models like CLIP, recent efforts have sought to harness their capabilities by either training a contrastive model from scratch with an extensive collection of region-label pairs or aligning the outputs of a detection model with image-level representations of region proposals. Despite notable progress, these approaches are plagued by computationally intensive training requirements, susceptibility to data noise, and deficiency in contextual information. To address these limitations, we explore the synergistic potential of off-the-shelf foundation models, leveraging their respective strengths in localization and semantics. We introduce a novel, generic, and efficient architecture, named RegionSpot, designed to integrate position-aware localization knowledge from a localization foundation model (e.g., SAM) with semantic information from a ViL model (e.g., CLIP). To fully exploit pretrained knowledge while minimizing training overhead, we keep both foundation models frozen, focusing optimization efforts solely on a lightweight attention-based knowledge integration module. Extensive experiments in open-world object recognition show that our RegionSpot achieves significant performance gain over prior alternatives, along with substantial computational savings (e.g., training our model with 3 million data in a single day using 8 V100 GPUs). RegionSpot outperforms GLIP-L by 2.9 in mAP on LVIS val set, with an even larger margin of 13.1 AP for more challenging and rare categories, and a 2.5 AP increase on ODinW. Furthermore, it exceeds GroundingDINO-L by 11.0 AP for rare categories on the LVIS minival set.
Authors: Sirui Cheng, Siyu Zhang, Jiayi Wu, Muchen Lan
Abstract: Within the multimodal field, large vision-language models (LVLMs) have made significant progress due to their strong perception and reasoning capabilities in the visual and language systems. However, LVLMs are still plagued by the two critical issues of object hallucination and factual accuracy, which limit the practicality of LVLMs in different scenarios. Furthermore, previous evaluation methods focus more on the comprehension and reasoning of language content but lack a comprehensive evaluation of multimodal interactions, thereby resulting in potential limitations. To this end, we propose a novel KNVQA-Eval, which is devoted to knowledge-based VQA task evaluation to reflect the factuality of multimodal LVLMs. To ensure the robustness and scalability of the evaluation, we develop a new KNVQA dataset by incorporating human judgment and perception, aiming to evaluate the accuracy of standard answers relative to AI-generated answers in knowledge-based VQA. This work not only comprehensively evaluates the contextual information of LVLMs using reliable human annotations, but also further analyzes the fine-grained capabilities of current methods to reveal potential avenues for subsequent optimization of LVLMs-based estimators. Our proposed VQA-Eval and corresponding dataset KNVQA will facilitate the development of automatic evaluation tools with the advantages of low cost, privacy protection, and reproducibility. Our code will be released upon publication.
Authors: Jiayun Luo, Siddhesh Khandelwal, Leonid Sigal, Boyang Li
Abstract: From image-text pairs, large-scale vision-language models (VLMs) learn to implicitly associate image regions with words, which prove effective for tasks like visual question answering. However, leveraging the learned association for open-vocabulary semantic segmentation remains a challenge. In this paper, we propose a simple, yet extremely effective, training-free technique, Plug-and-Play Open-Vocabulary Semantic Segmentation (PnP-OVSS) for this task. PnP-OVSS leverages a VLM with direct text-to-image cross-attention and an image-text matching loss. To balance between over-segmentation and under-segmentation, we introduce Salience Dropout; by iteratively dropping patches that the model is most attentive to, we are able to better resolve the entire extent of the segmentation mask. PnP-OVSS does not require any neural network training and performs hyperparameter tuning without the need for any segmentation annotations, even for a validation set. PnP-OVSS demonstrates substantial improvements over comparable baselines (+26.2% mIoU on Pascal VOC, +20.5% mIoU on MS COCO, +3.1% mIoU on COCO Stuff and +3.0% mIoU on ADE20K). Our codebase is at https://github.com/letitiabanana/PnP-OVSS.
Authors: Li Hu, Xin Gao, Peng Zhang, Ke Sun, Bang Zhang, Liefeng Bo
Abstract: Character Animation aims to generating character videos from still images through driving signals. Currently, diffusion models have become the mainstream in visual generation research, owing to their robust generative capabilities. However, challenges persist in the realm of image-to-video, especially in character animation, where temporally maintaining consistency with detailed information from character remains a formidable problem. In this paper, we leverage the power of diffusion models and propose a novel framework tailored for character animation. To preserve consistency of intricate appearance features from reference image, we design ReferenceNet to merge detail features via spatial attention. To ensure controllability and continuity, we introduce an efficient pose guider to direct character's movements and employ an effective temporal modeling approach to ensure smooth inter-frame transitions between video frames. By expanding the training data, our approach can animate arbitrary characters, yielding superior results in character animation compared to other image-to-video methods. Furthermore, we evaluate our method on benchmarks for fashion video and human dance synthesis, achieving state-of-the-art results.
Authors: Jaeseok Byun, Dohoon Kim, Taesup Moon
Abstract: We consider a critical issue of false negatives in Vision-Language Pre-training (VLP), a challenge that arises from the inherent many-to-many correspondence of image-text pairs in large-scale web-crawled datasets. The presence of false negatives can impede achieving optimal performance and even lead to a significant performance drop. To address this challenge, we propose MAFA (MAnaging FAlse negatives), which consists of two pivotal components building upon the recently developed GRouped mIni-baTch sampling (GRIT) strategy: 1) an efficient connection mining process that identifies and converts false negatives into positives, and 2) label smoothing for the image-text contrastive (ITC) loss. Our comprehensive experiments verify the effectiveness of MAFA across multiple downstream tasks, emphasizing the crucial role of addressing false negatives in VLP, potentially even surpassing the importance of addressing false positives. In addition, the compatibility of MAFA with the recent BLIP-family model is also demonstrated. Code is available at https://github.com/jaeseokbyun/MAFA.
Authors: Tuan Truong, Farnaz Khun Jush, Matthias Lenga
Abstract: Near- and duplicate image detection is a critical concern in the field of medical imaging. Medical datasets often contain similar or duplicate images from various sources, which can lead to significant performance issues and evaluation biases, especially in machine learning tasks due to data leakage between training and testing subsets. In this paper, we present an approach for identifying near- and duplicate 3D medical images leveraging publicly available 2D computer vision embeddings. We assessed our approach by comparing embeddings extracted from two state-of-the-art self-supervised pretrained models and two different vector index structures for similarity retrieval. We generate an experimental benchmark based on the publicly available Medical Segmentation Decathlon dataset. The proposed method yields promising results for near- and duplicate image detection achieving a mean sensitivity and specificity of 0.9645 and 0.8559, respectively.
Authors: Zike Wu, Pan Zhou, Xuanyu Yi, Xiaoding Yuan, Hanwang Zhang
Abstract: Score distillation sampling (SDS) and its variants have greatly boosted the development of text-to-3D generation, but are vulnerable to geometry collapse and poor textures yet. To solve this issue, we first deeply analyze the SDS and find that its distillation sampling process indeed corresponds to the trajectory sampling of a stochastic differential equation (SDE): SDS samples along an SDE trajectory to yield a less noisy sample which then serves as a guidance to optimize a 3D model. However, the randomness in SDE sampling often leads to a diverse and unpredictable sample which is not always less noisy, and thus is not a consistently correct guidance, explaining the vulnerability of SDS. Since for any SDE, there always exists an ordinary differential equation (ODE) whose trajectory sampling can deterministically and consistently converge to the desired target point as the SDE, we propose a novel and effective "Consistent3D" method that explores the ODE deterministic sampling prior for text-to-3D generation. Specifically, at each training iteration, given a rendered image by a 3D model, we first estimate its desired 3D score function by a pre-trained 2D diffusion model, and build an ODE for trajectory sampling. Next, we design a consistency distillation sampling loss which samples along the ODE trajectory to generate two adjacent samples and uses the less noisy sample to guide another more noisy one for distilling the deterministic prior into the 3D model. Experimental results show the efficacy of our Consistent3D in generating high-fidelity and diverse 3D objects and large-scale scenes, as shown in Fig. 1. The codes are available at https://github.com/sail-sg/Consistent3D.
Authors: Tosin Adewumi, Nudrat Habib, Lama Alkhaled, Elisa Barney
Abstract: We introduce Instruction Document Visual Question Answering (iDocVQA) dataset and Large Language Document (LLaDoc) model, for training Language-Vision (LV) models for document analysis and predictions on document images, respectively. Usually, deep neural networks for the DocVQA task are trained on datasets lacking instructions. We show that using instruction-following datasets improves performance. We compare performance across document-related datasets using the recent state-of-the-art (SotA) Large Language and Vision Assistant (LLaVA)1.5 as the base model. We also evaluate the performance of the derived models for object hallucination using the Polling-based Object Probing Evaluation (POPE) dataset. The results show that instruction-tuning performance ranges from 11X to 32X of zero-shot performance and from 0.1% to 4.2% over non-instruction (traditional task) finetuning. Despite the gains, these still fall short of human performance (94.36%), implying there's much room for improvement.
Authors: Shoubin Yu, Jaehong Yoon, Mohit Bansal
Abstract: Despite impressive advancements in recent multimodal reasoning approaches, they are still limited in flexibility and efficiency, as these models typically process only a few fixed modality inputs and require updates to numerous parameters. This paper tackles these critical challenges and proposes CREMA, a generalizable, highly efficient, and modular modality-fusion framework that can incorporate any new modality to enhance video reasoning. We first augment multiple informative modalities (such as optical flow, 3D point cloud, audio, thermal heatmap, and touch map) from given videos without extra human annotation by leveraging sensors or existing pre-trained models. Next, we introduce a query transformer with multiple parameter-efficient modules associated with each accessible modality. It projects diverse modality features to the LLM token embedding space, allowing the model to integrate different data types for response generation. Furthermore, we propose a novel progressive multimodal fusion design supported by a lightweight fusion module and modality-sequential training strategy. It helps compress information across various assisting modalities, maintaining computational efficiency in the LLM while improving performance. We validate our method on 7 video-language reasoning tasks assisted by diverse modalities, including conventional VideoQA and Video-Audio/3D/Touch/Thermal QA, and achieve better/equivalent performance against strong multimodal LLMs, including OneLLM, BLIP-2, and SeViLA while reducing over 90% trainable parameters. We provide extensive analyses of CREMA, including the impact of each modality on reasoning domains, the design of the fusion module, and example visualizations.
Authors: Matteo Paiano, Stefano Martina, Carlotta Giannelli, Filippo Caruso
Abstract: The availability of data is limited in some fields, especially for object detection tasks, where it is necessary to have correctly labeled bounding boxes around each object. A notable example of such data scarcity is found in the domain of marine biology, where it is useful to develop methods to automatically detect submarine species for environmental monitoring. To address this data limitation, the state-of-the-art machine learning strategies employ two main approaches. The first involves pretraining models on existing datasets before generalizing to the specific domain of interest. The second strategy is to create synthetic datasets specifically tailored to the target domain using methods like copy-paste techniques or ad-hoc simulators. The first strategy often faces a significant domain shift, while the second demands custom solutions crafted for the specific task. In response to these challenges, here we propose a transfer learning framework that is valid for a generic scenario. In this framework, generated images help to improve the performances of an object detector in a few-real data regime. This is achieved through a diffusion-based generative model that was pretrained on large generic datasets. With respect to the state-of-the-art, we find that it is not necessary to fine tune the generative model on the specific domain of interest. We believe that this is an important advance because it mitigates the labor-intensive task of manual labeling the images in object detection tasks. We validate our approach focusing on fishes in an underwater environment, and on the more common domain of cars in an urban setting. Our method achieves detection performance comparable to models trained on thousands of images, using only a few hundreds of input data. Our results pave the way for new generative AI-based protocols for machine learning applications in various domains.
Authors: Muyang He, Yexin Liu, Boya Wu, Jianhao Yuan, Yueze Wang, Tiejun Huang, Bo Zhao
Abstract: Multimodal Large Language Models (MLLMs) have demonstrated notable capabilities in general visual understanding and reasoning tasks. However, their deployment is hindered by substantial computational costs in both training and inference, limiting accessibility to the broader research and user communities. A straightforward solution is to leverage smaller pre-trained vision and language models, which inevitably cause significant performance drops. In this paper, we demonstrate the possibility of training a smaller but better MLLM with high-quality training data. Specifically, we introduce Bunny, a family of lightweight MLLMs with flexible vision and language backbones for efficient multimodal learning from selected training data. Experiments show that our Bunny-4B/8B outperforms the state-of-the-art large MLLMs on multiple benchmarks. We expect that this work can provide the community with a clean and flexible open-source tool for further research and development. The code, models, and data can be found in https://github.com/BAAI-DCAI/Bunny.
Authors: Wei Zhang, Miaoxin Cai, Tong Zhang, Guoqiang Lei, Yin Zhuang, Xuerui Mao
Abstract: Ship detection needs to identify ship locations from remote sensing (RS) scenes. Due to different imaging payloads, various appearances of ships, and complicated background interference from the bird's eye view, it is difficult to set up a unified paradigm for achieving multi-source ship detection. To address this challenge, in this article, leveraging the large language models (LLMs)'s powerful generalization ability, a unified visual-language model called Popeye is proposed for multi-source ship detection from RS imagery. Specifically, to bridge the interpretation gap between the multi-source images for ship detection, a novel unified labeling paradigm is designed to integrate different visual modalities and the various ship detection ways, i.e., horizontal bounding box (HBB) and oriented bounding box (OBB). Subsequently, the hybrid experts encoder is designed to refine multi-scale visual features, thereby enhancing visual perception. Then, a visual-language alignment method is developed for Popeye to enhance interactive comprehension ability between visual and language content. Furthermore, an instruction adaption mechanism is proposed for transferring the pre-trained visual-language knowledge from the nature scene into the RS domain for multi-source ship detection. In addition, the segment anything model (SAM) is also seamlessly integrated into the proposed Popeye to achieve pixel-level ship segmentation without additional training costs. Finally, extensive experiments are conducted on the newly constructed ship instruction dataset named MMShip, and the results indicate that the proposed Popeye outperforms current specialist, open-vocabulary, and other visual-language models for zero-shot multi-source ship detection.
Authors: Yulong Liu, Yongqiang Ma, Guibo Zhu, Haodong Jing, Nanning Zheng
Abstract: Deciphering visual content from functional Magnetic Resonance Imaging (fMRI) helps illuminate the human vision system. However, the scarcity of fMRI data and noise hamper brain decoding model performance. Previous approaches primarily employ subject-specific models, sensitive to training sample size. In this paper, we explore a straightforward but overlooked solution to address data scarcity. We propose shallow subject-specific adapters to map cross-subject fMRI data into unified representations. Subsequently, a shared deeper decoding model decodes cross-subject features into the target feature space. During training, we leverage both visual and textual supervision for multi-modal brain decoding. Our model integrates a high-level perception decoding pipeline and a pixel-wise reconstruction pipeline guided by high-level perceptions, simulating bottom-up and top-down processes in neuroscience. Empirical experiments demonstrate robust neural representation learning across subjects for both pipelines. Moreover, merging high-level and low-level information improves both low-level and high-level reconstruction metrics. Additionally, we successfully transfer learned general knowledge to new subjects by training new adapters with limited training data. Compared to previous state-of-the-art methods, notably pre-training-based methods (Mind-Vis and fMRI-PTE), our approach achieves comparable or superior results across diverse tasks, showing promise as an alternative method for cross-subject fMRI data pre-training. Our code and pre-trained weights will be publicly released at https://github.com/YulongBonjour/See_Through_Their_Minds.
URLs: https://github.com/YulongBonjour/See_Through_Their_Minds.
Authors: Fangqiang Ding, Lawrence Zhu, Xiangyu Wen, Gaowen Liu, Chris Xiaoxuan Lu
Abstract: In this work, we present ThermoHands, a new benchmark for thermal image-based egocentric 3D hand pose estimation, aimed at overcoming challenges like varying lighting conditions and obstructions (e.g., handwear). The benchmark includes a multi-view and multi-spectral dataset collected from 28 subjects performing hand-object and hand-virtual interactions under diverse scenarios, accurately annotated with 3D hand poses through an automated process. We introduce a new baseline method, TherFormer, utilizing dual transformer modules for effective egocentric 3D hand pose estimation in thermal imagery. Our experimental results highlight TherFormer's leading performance and affirm thermal imaging's effectiveness in enabling robust 3D hand pose estimation in adverse conditions.
Authors: Zhenglin Li, Yangchen Huang, Mengran Zhu, Jingyu Zhang, JingHao Chang, Houze Liu
Abstract: Change detection is a fundamental task in computer vision that processes a bi-temporal image pair to differentiate between semantically altered and unaltered regions. Large language models (LLMs) have been utilized in various domains for their exceptional feature extraction capabilities and have shown promise in numerous downstream applications. In this study, we harness the power of a pre-trained LLM, extracting feature maps from extensive datasets, and employ an auxiliary network to detect changes. Unlike existing LLM-based change detection methods that solely focus on deriving high-quality feature maps, our approach emphasizes the manipulation of these feature maps to enhance semantic relevance.
Authors: Maxence Bideaux, Alice Phe, Mohamed Chaouch, Bertrand Luvison, Quoc-Cuong Pham
Abstract: We introduce 3D-COCO, an extension of the original MS-COCO dataset providing 3D models and 2D-3D alignment annotations. 3D-COCO was designed to achieve computer vision tasks such as 3D reconstruction or image detection configurable with textual, 2D image, and 3D CAD model queries. We complete the existing MS-COCO dataset with 28K 3D models collected on ShapeNet and Objaverse. By using an IoU-based method, we match each MS-COCO annotation with the best 3D models to provide a 2D-3D alignment. The open-source nature of 3D-COCO is a premiere that should pave the way for new research on 3D-related topics. The dataset and its source codes is available at https://kalisteo.cea.fr/index.php/coco3d-object-detection-and-reconstruction/
URLs: https://kalisteo.cea.fr/index.php/coco3d-object-detection-and-reconstruction/
Authors: Ming-Kun Xie, Jia-Hao Xiao, Pei Peng, Gang Niu, Masashi Sugiyama, Sheng-Jun Huang
Abstract: The key to multi-label image classification (MLC) is to improve model performance by leveraging label correlations. Unfortunately, it has been shown that overemphasizing co-occurrence relationships can cause the overfitting issue of the model, ultimately leading to performance degradation. In this paper, we provide a causal inference framework to show that the correlative features caused by the target object and its co-occurring objects can be regarded as a mediator, which has both positive and negative impacts on model predictions. On the positive side, the mediator enhances the recognition performance of the model by capturing co-occurrence relationships; on the negative side, it has the harmful causal effect that causes the model to make an incorrect prediction for the target object, even when only co-occurring objects are present in an image. To address this problem, we propose a counterfactual reasoning method to measure the total direct effect, achieved by enhancing the direct effect caused only by the target object. Due to the unknown location of the target object, we propose patching-based training and inference to accomplish this goal, which divides an image into multiple patches and identifies the pivot patch that contains the target object. Experimental results on multiple benchmark datasets with diverse configurations validate that the proposed method can achieve state-of-the-art performance.
Authors: Yongquan Qu, Juan Nathaniel, Shuolin Li, Pierre Gentine
Abstract: Robust integration of physical knowledge and data is key to improve computational simulations, such as Earth system models. Data assimilation is crucial for achieving this goal because it provides a systematic framework to calibrate model outputs with observations, which can include remote sensing imagery and ground station measurements, with uncertainty quantification. Conventional methods, including Kalman filters and variational approaches, inherently rely on simplifying linear and Gaussian assumptions, and can be computationally expensive. Nevertheless, with the rapid adoption of data-driven methods in many areas of computational sciences, we see the potential of emulating traditional data assimilation with deep learning, especially generative models. In particular, the diffusion-based probabilistic framework has large overlaps with data assimilation principles: both allows for conditional generation of samples with a Bayesian inverse framework. These models have shown remarkable success in text-conditioned image generation or image-controlled video synthesis. Likewise, one can frame data assimilation as observation-conditioned state calibration. In this work, we propose SLAMS: Score-based Latent Assimilation in Multimodal Setting. Specifically, we assimilate in-situ weather station data and ex-situ satellite imagery to calibrate the vertical temperature profiles, globally. Through extensive ablation, we demonstrate that SLAMS is robust even in low-resolution, noisy, and sparse data settings. To our knowledge, our work is the first to apply deep generative framework for multimodal data assimilation using real-world datasets; an important step for building robust computational simulators, including the next-generation Earth system models. Our code is available at: https://github.com/yongquan-qu/SLAMS
Authors: Ivica Obadic, Alex Levering, Lars Pennig, Dario Oliveira, Diego Marcos, Xiaoxiang Zhu
Abstract: Predicting socioeconomic indicators from satellite imagery with deep learning has become an increasingly popular research direction. Post-hoc concept-based explanations can be an important step towards broader adoption of these models in policy-making as they enable the interpretation of socioeconomic outcomes based on visual concepts that are intuitive to humans. In this paper, we study the interplay between representation learning using an additional task-specific contrastive loss and post-hoc concept explainability for socioeconomic studies. Our results on two different geographical locations and tasks indicate that the task-specific pretraining imposes a continuous ordering of the latent space embeddings according to the socioeconomic outcomes. This improves the model's interpretability as it enables the latent space of the model to associate concepts encoding typical urban and natural area patterns with continuous intervals of socioeconomic outcomes. Further, we illustrate how analyzing the model's conceptual sensitivity for the intervals of socioeconomic outcomes can shed light on new insights for urban studies.
Authors: Kiyoon Jeong, Woojun Lee, Woongchan Nam, Minjeong Ma, Pilsung Kang
Abstract: This report presents the ECO (Ensembled Clip score and cOnsensus score) pipeline from team DSBA LAB, which is a new framework used to evaluate and rank captions for a given image. ECO selects the most accurate caption describing image. It is made possible by combining an Ensembled CLIP score, which considers the semantic alignment between the image and captions, with a Consensus score that accounts for the essentialness of the captions. Using this framework, we achieved notable success in the CVPR 2024 Workshop Challenge on Caption Re-ranking Evaluation at the New Frontiers for Zero-Shot Image Captioning Evaluation (NICE). Specifically, we secured third place based on the CIDEr metric, second in both the SPICE and METEOR metrics, and first in the ROUGE-L and all BLEU Score metrics. The code and configuration for the ECO framework are available at https://github.com/DSBA-Lab/ECO .
Authors: Emmanuelle Bourigault, Pauline Bourigault
Abstract: Generating consistent multiple views for 3D reconstruction tasks is still a challenge to existing image-to-3D diffusion models. Generally, incorporating 3D representations into diffusion model decrease the model's speed as well as generalizability and quality. This paper proposes a general framework to generate consistent multi-view images from single image or leveraging scene representation transformer and view-conditioned diffusion model. In the model, we introduce epipolar geometry constraints and multi-view attention to enforce 3D consistency. From as few as one image input, our model is able to generate 3D meshes surpassing baselines methods in evaluation metrics, including PSNR, SSIM and LPIPS.
Authors: Fakrul Islam Tushar, Avivah Wang, Lavsen Dahal, Michael R. Harowicz, Kyle J. Lafata, Tina D. Tailor, Joseph Y. Lo
Abstract: Lung cancer's high mortality rate can be mitigated by early detection, increasingly reliant on AI for diagnostic imaging. However, AI model performance depends on training and validation datasets. This study develops and validates AI models for both nodule detection and cancer classification tasks. For detection, two models (DLCSD-mD and LUNA16-mD) were developed using the Duke Lung Cancer Screening Dataset (DLCSD), with over 2,000 CT scans from 1,613 patients and more than 3,000 annotations. These models were evaluated on internal (DLCSD) and external datasets, including LUNA16 (601 patients, 1186 nodules) and NLST (969 patients, 1192 nodules), using FROC analysis and AUC metrics. For classification, five models were developed and tested: a randomly initialized 3D ResNet50, Genesis, MedNet3D, an enhanced ResNet50 using Strategic Warm-Start++ (SWS++), and a linear classifier analyzing features from the Foundation Model for Cancer Biomarkers (FMCB). These models were trained to distinguish between benign and malignant nodules and evaluated using AUC analysis on internal (DLCSD) and external datasets, including LUNA16 (433 patients, 677 nodules) and NLST. The DLCSD-mD model achieved an AUC of 0.93 (95% CI: 0.91-0.94) on the internal DLCSD dataset. External validation results were 0.97 (95% CI: 0.96-0.98) on LUNA16 and 0.75 (95% CI: 0.73-0.76) on NLST. For classification, the ResNet50-SWS++ model recorded AUCs of 0.71 (95% CI: 0.61-0.81) on DLCSD, 0.90 (95% CI: 0.87-0.93) on LUNA16, and 0.81 (95% CI: 0.79-0.82) on NLST. Other models showed varying performance across datasets, underscoring the importance of diverse model approaches. This benchmarking establishes DLCSD as a reliable resource for lung cancer AI research.
Authors: Peng Gao, Le Zhuo, Dongyang Liu, Ruoyi Du, Xu Luo, Longtian Qiu, Yuhang Zhang, Chen Lin, Rongjie Huang, Shijie Geng, Renrui Zhang, Junlin Xi, Wenqi Shao, Zhengkai Jiang, Tianshuo Yang, Weicai Ye, He Tong, Jingwen He, Yu Qiao, Hongsheng Li
Abstract: Sora unveils the potential of scaling Diffusion Transformer for generating photorealistic images and videos at arbitrary resolutions, aspect ratios, and durations, yet it still lacks sufficient implementation details. In this technical report, we introduce the Lumina-T2X family - a series of Flow-based Large Diffusion Transformers (Flag-DiT) equipped with zero-initialized attention, as a unified framework designed to transform noise into images, videos, multi-view 3D objects, and audio clips conditioned on text instructions. By tokenizing the latent spatial-temporal space and incorporating learnable placeholders such as [nextline] and [nextframe] tokens, Lumina-T2X seamlessly unifies the representations of different modalities across various spatial-temporal resolutions. This unified approach enables training within a single framework for different modalities and allows for flexible generation of multimodal data at any resolution, aspect ratio, and length during inference. Advanced techniques like RoPE, RMSNorm, and flow matching enhance the stability, flexibility, and scalability of Flag-DiT, enabling models of Lumina-T2X to scale up to 7 billion parameters and extend the context window to 128K tokens. This is particularly beneficial for creating ultra-high-definition images with our Lumina-T2I model and long 720p videos with our Lumina-T2V model. Remarkably, Lumina-T2I, powered by a 5-billion-parameter Flag-DiT, requires only 35% of the training computational costs of a 600-million-parameter naive DiT. Our further comprehensive analysis underscores Lumina-T2X's preliminary capability in resolution extrapolation, high-resolution editing, generating consistent 3D views, and synthesizing videos with seamless transitions. We expect that the open-sourcing of Lumina-T2X will further foster creativity, transparency, and diversity in the generative AI community.
Authors: Chengcheng Feng, Mu He, Qiuyu Tian, Haojie Yin, Xiaofang Zhao, Hongwei Tang, Xingqiang Wei
Abstract: As deep learning technology continues to advance, image generation models, especially models like Stable Diffusion, are finding increasingly widespread application in visual arts creation. However, these models often face challenges such as overfitting, lack of stability in generated results, and difficulties in accurately capturing the features desired by creators during the fine-tuning process. In response to these challenges, we propose an innovative method that integrates Singular Value Decomposition (SVD) into the Low-Rank Adaptation (LoRA) parameter update strategy, aimed at enhancing the fine-tuning efficiency and output quality of image generation models. By incorporating SVD within the LoRA framework, our method not only effectively reduces the risk of overfitting but also enhances the stability of model outputs, and captures subtle, creator-desired feature adjustments more accurately. We evaluated our method on multiple datasets, and the results show that, compared to traditional fine-tuning methods, our approach significantly improves the model's generalization ability and creative flexibility while maintaining the quality of generation. Moreover, this method maintains LoRA's excellent performance under resource-constrained conditions, allowing for significant improvements in image generation quality without sacrificing the original efficiency and resource advantages.
Authors: Fangqiang Ding, Xiangyu Wen, Lawrence Zhu, Yiming Li, Chris Xiaoxuan Lu
Abstract: 3D occupancy-based perception pipeline has significantly advanced autonomous driving by capturing detailed scene descriptions and demonstrating strong generalizability across various object categories and shapes. Current methods predominantly rely on LiDAR or camera inputs for 3D occupancy prediction. These methods are susceptible to adverse weather conditions, limiting the all-weather deployment of self-driving cars. To improve perception robustness, we leverage the recent advances in automotive radars and introduce a novel approach that utilizes 4D imaging radar sensors for 3D occupancy prediction. Our method, RadarOcc, circumvents the limitations of sparse radar point clouds by directly processing the 4D radar tensor, thus preserving essential scene details. RadarOcc innovatively addresses the challenges associated with the voluminous and noisy 4D radar data by employing Doppler bins descriptors, sidelobe-aware spatial sparsification, and range-wise self-attention mechanisms. To minimize the interpolation errors associated with direct coordinate transformations, we also devise a spherical-based feature encoding followed by spherical-to-Cartesian feature aggregation. We benchmark various baseline methods based on distinct modalities on the public K-Radar dataset. The results demonstrate RadarOcc's state-of-the-art performance in radar-based 3D occupancy prediction and promising results even when compared with LiDAR- or camera-based methods. Additionally, we present qualitative evidence of the superior performance of 4D radar in adverse weather conditions and explore the impact of key pipeline components through ablation studies.
Authors: Artem Lukoianov, Haitz S\'aez de Oc\'ariz Borde, Kristjan Greenewald, Vitor Campagnolo Guizilini, Timur Bagautdinov, Vincent Sitzmann, Justin Solomon
Abstract: While 2D diffusion models generate realistic, high-detail images, 3D shape generation methods like Score Distillation Sampling (SDS) built on these 2D diffusion models produce cartoon-like, over-smoothed shapes. To help explain this discrepancy, we show that the image guidance used in Score Distillation can be understood as the velocity field of a 2D denoising generative process, up to the choice of a noise term. In particular, after a change of variables, SDS resembles a high-variance version of Denoising Diffusion Implicit Models (DDIM) with a differently-sampled noise term: SDS introduces noise i.i.d. randomly at each step, while DDIM infers it from the previous noise predictions. This excessive variance can lead to over-smoothing and unrealistic outputs. We show that a better noise approximation can be recovered by inverting DDIM in each SDS update step. This modification makes SDS's generative process for 2D images almost identical to DDIM. In 3D, it removes over-smoothing, preserves higher-frequency detail, and brings the generation quality closer to that of 2D samplers. Experimentally, our method achieves better or similar 3D generation quality compared to other state-of-the-art Score Distillation methods, all without training additional neural networks or multi-view supervision, and providing useful insights into relationship between 2D and 3D asset generation with diffusion models.
Authors: Kailu Wu, Fangfu Liu, Zhihan Cai, Runjie Yan, Hanyang Wang, Yating Hu, Yueqi Duan, Kaisheng Ma
Abstract: In this work, we introduce Unique3D, a novel image-to-3D framework for efficiently generating high-quality 3D meshes from single-view images, featuring state-of-the-art generation fidelity and strong generalizability. Previous methods based on Score Distillation Sampling (SDS) can produce diversified 3D results by distilling 3D knowledge from large 2D diffusion models, but they usually suffer from long per-case optimization time with inconsistent issues. Recent works address the problem and generate better 3D results either by finetuning a multi-view diffusion model or training a fast feed-forward model. However, they still lack intricate textures and complex geometries due to inconsistency and limited generated resolution. To simultaneously achieve high fidelity, consistency, and efficiency in single image-to-3D, we propose a novel framework Unique3D that includes a multi-view diffusion model with a corresponding normal diffusion model to generate multi-view images with their normal maps, a multi-level upscale process to progressively improve the resolution of generated orthographic multi-views, as well as an instant and consistent mesh reconstruction algorithm called ISOMER, which fully integrates the color and geometric priors into mesh results. Extensive experiments demonstrate that our Unique3D significantly outperforms other image-to-3D baselines in terms of geometric and textural details.
Authors: Jingyun Xue, Hongfa Wang, Qi Tian, Yue Ma, Andong Wang, Zhiyuan Zhao, Shaobo Min, Wenzhe Zhao, Kaihao Zhang, Heung-Yeung Shum, Wei Liu, Mengyang Liu, Wenhan Luo
Abstract: Pose-controllable character video generation is in high demand with extensive applications for fields such as automatic advertising and content creation on social media platforms. While existing character image animation methods using pose sequences and reference images have shown promising performance, they tend to struggle with incoherent animation in complex scenarios, such as multiple character animation and body occlusion. Additionally, current methods request large-scale high-quality videos with stable backgrounds and temporal consistency as training datasets, otherwise, their performance will greatly deteriorate. These two issues hinder the practical utilization of character image animation tools. In this paper, we propose a practical and robust framework Follow-Your-Pose v2, which can be trained on noisy open-sourced videos readily available on the internet. Multi-condition guiders are designed to address the challenges of background stability, body occlusion in multi-character generation, and consistency of character appearance. Moreover, to fill the gap of fair evaluation of multi-character pose animation, we propose a new benchmark comprising approximately 4,000 frames. Extensive experiments demonstrate that our approach outperforms state-of-the-art methods by a margin of over 35% across 2 datasets and on 7 metrics. Meanwhile, qualitative assessments reveal a significant improvement in the quality of generated video, particularly in scenarios involving complex backgrounds and body occlusion of multi-character, suggesting the superiority of our approach.
Authors: Haosen Yang, Chenhao Zhang, Wenqing Wang, Marco Volino, Adrian Hilton, Li Zhang, Xiatian Zhu
Abstract: Point management is a critical component in optimizing 3D Gaussian Splatting (3DGS) models, as the point initiation (e.g., via structure from motion) is distributionally inappropriate. Typically, the Adaptive Density Control (ADC) algorithm is applied, leveraging view-averaged gradient magnitude thresholding for point densification, opacity thresholding for pruning, and regular all-points opacity reset. However, we reveal that this strategy is limited in tackling intricate/special image regions (e.g., transparent) as it is unable to identify all the 3D zones that require point densification, and lacking an appropriate mechanism to handle the ill-conditioned points with negative impacts (occlusion due to false high opacity). To address these limitations, we propose a Localized Point Management (LPM) strategy, capable of identifying those error-contributing zones in the highest demand for both point addition and geometry calibration. Zone identification is achieved by leveraging the underlying multiview geometry constraints, with the guidance of image rendering errors. We apply point densification in the identified zone, whilst resetting the opacity of those points residing in front of these regions so that a new opportunity is created to correct ill-conditioned points. Serving as a versatile plugin, LPM can be seamlessly integrated into existing 3D Gaussian Splatting models. Experimental evaluation across both static 3D and dynamic 4D scenes validate the efficacy of our LPM strategy in boosting a variety of existing 3DGS models both quantitatively and qualitatively. Notably, LPM improves both vanilla 3DGS and SpaceTimeGS to achieve state-of-the-art rendering quality while retaining real-time speeds, outperforming on challenging datasets such as Tanks & Temples and the Neural 3D Video Dataset.
Authors: Sergio Casas, Ben Agro, Jiageng Mao, Thomas Gilles, Alexander Cui, Thomas Li, Raquel Urtasun
Abstract: The tasks of object detection and trajectory forecasting play a crucial role in understanding the scene for autonomous driving. These tasks are typically executed in a cascading manner, making them prone to compounding errors. Furthermore, there is usually a very thin interface between the two tasks, creating a lossy information bottleneck. To address these challenges, our approach formulates the union of the two tasks as a trajectory refinement problem, where the first pose is the detection (current time), and the subsequent poses are the waypoints of the multiple forecasts (future time). To tackle this unified task, we design a refinement transformer that infers the presence, pose, and multi-modal future behaviors of objects directly from LiDAR point clouds and high-definition maps. We call this model DeTra, short for object Detection and Trajectory forecasting. In our experiments, we observe that \ourmodel{} outperforms the state-of-the-art on Argoverse 2 Sensor and Waymo Open Dataset by a large margin, across a broad range of metrics. Last but not least, we perform extensive ablation studies that show the value of refinement for this task, that every proposed component contributes positively to its performance, and that key design choices were made.
Authors: Haokun Zhou, Yipeng Hong
Abstract: This study assesses the ability of Large Vision-Language Models (LVLMs) to differentiate between AI-generated and human-generated images. It introduces a new automated benchmark construction method for this evaluation. The experiment compared common LVLMs with human participants using a mixed dataset of AI and human-created images. Results showed that LVLMs could distinguish between the image types to some extent but exhibited a rightward bias, and perform significantly worse compared to humans. To build on these findings, we developed an automated benchmark construction process using AI. This process involved topic retrieval, narrative script generation, error embedding, and image generation, creating a diverse set of text-image pairs with intentional errors. We validated our method through constructing two caparable benchmarks. This study highlights the strengths and weaknesses of LVLMs in real-world understanding and advances benchmark construction techniques, providing a scalable and automatic approach for AI model evaluation.
Authors: Shaurya Dewan, Rushikesh Zawar, Prakanshul Saxena, Yingshan Chang, Andrew Luo, Yonatan Bisk
Abstract: Text-to-image diffusion models have made significant progress in generating naturalistic images from textual inputs, and demonstrate the capacity to learn and represent complex visual-semantic relationships. While these diffusion models have achieved remarkable success, the underlying mechanisms driving their performance are not yet fully accounted for, with many unanswered questions surrounding what they learn, how they represent visual-semantic relationships, and why they sometimes fail to generalize. Our work presents Diffusion Partial Information Decomposition (DiffusionPID), a novel technique that applies information-theoretic principles to decompose the input text prompt into its elementary components, enabling a detailed examination of how individual tokens and their interactions shape the generated image. We introduce a formal approach to analyze the uniqueness, redundancy, and synergy terms by applying PID to the denoising model at both the image and pixel level. This approach enables us to characterize how individual tokens and their interactions affect the model output. We first present a fine-grained analysis of characteristics utilized by the model to uniquely localize specific concepts, we then apply our approach in bias analysis and show it can recover gender and ethnicity biases. Finally, we use our method to visually characterize word ambiguity and similarity from the model's perspective and illustrate the efficacy of our method for prompt intervention. Our results show that PID is a potent tool for evaluating and diagnosing text-to-image diffusion models.
Authors: Khiem Le, Zhichun Guo, Kaiwen Dong, Xiaobao Huang, Bozhao Nan, Roshni Iyer, Xiangliang Zhang, Olaf Wiest, Wei Wang, Nitesh V. Chawla
Abstract: Recently, Large Language Models (LLMs) with their strong task-handling capabilities have shown remarkable advancements across a spectrum of fields, moving beyond natural language understanding. However, their proficiency within the chemistry domain remains restricted, especially in solving professional molecule-related tasks. This challenge is attributed to their inherent limitations in comprehending molecules using only common textual representations, i.e., SMILES strings. In this study, we seek to enhance the ability of LLMs to comprehend molecules by designing and equipping them with a multi-modal external module, namely MolX. In particular, instead of directly using a SMILES string to represent a molecule, we utilize specific encoders to extract fine-grained features from both SMILES string and 2D molecular graph representations for feeding into an LLM. Moreover, a human-defined molecular fingerprint is incorporated to leverage its embedded domain knowledge. Then, to establish an alignment between MolX and the LLM's textual input space, the whole model in which the LLM is frozen, is pre-trained with a versatile strategy including a diverse set of tasks. Extensive experimental evaluations demonstrate that our proposed method only introduces a small number of trainable parameters while outperforming baselines on various downstream molecule-related tasks ranging from molecule-to-text translation to retrosynthesis, with and without fine-tuning the LLM.
Authors: Weiqing Xiao, Wei Zhao
Abstract: Due to the difficulty in obtaining real samples and ground truth, the generalization performance and the fine-tuned performance are critical for the feasibility of stereo matching methods in real-world applications. However, the presence of substantial disparity distributions and density variations across different datasets presents significant challenges for the generalization and fine-tuning of the model. In this paper, we propose a novel stereo matching method, called SR-Stereo, which mitigates the distributional differences across different datasets by predicting the disparity clips and uses a loss weight related to the regression target scale to improve the accuracy of the disparity clips. Moreover, this stepwise regression architecture can be easily extended to existing iteration-based methods to improve the performance without changing the structure. In addition, to mitigate the edge blurring of the fine-tuned model on sparse ground truth, we propose Domain Adaptation Based on Pre-trained Edges (DAPE). Specifically, we use the predicted disparity and RGB image to estimate the edge map of the target domain image. The edge map is filtered to generate edge map background pseudo-labels, which together with the sparse ground truth disparity on the target domain are used as a supervision to jointly fine-tune the pre-trained stereo matching model. These proposed methods are extensively evaluated on SceneFlow, KITTI, Middbury 2014 and ETH3D. The SR-Stereo achieves competitive disparity estimation performance and state-of-the-art cross-domain generalisation performance. Meanwhile, the proposed DAPE significantly improves the disparity estimation performance of fine-tuned models, especially in the textureless and detail regions.
Authors: Che Liu, Zhongwei Wan, Yuqi Wang, Hui Shen, Haozhe Wang, Kangyu Zheng, Mi Zhang, Rossella Arcucci
Abstract: Automatic radiology report generation can significantly benefit the labor-intensive process of report writing by radiologists, especially for 3D radiographs like CT scans, which are crucial for broad clinical diagnostics yet underexplored compared to 2D radiographs. Existing methods often handle 3D volumes either slice-wise or with aggressive downsampling due to current GPU memory limitations, which results in a loss of the inherent 3D nature and critical details. To overcome these issues, we introduce a novel framework that efficiently and effectively generates radiology reports for high-resolution (HR) 3D volumes, based on large language models (LLMs). Specifically, our framework utilizes low-resolution (LR) visual tokens as queries to mine information from HR tokens, preserving detailed HR information while reducing computational costs by only processing HR informed LR visual queries. Further benefiting the field, we curate and release BIMCV-RG, a new dataset with 5,328 HR 3D volumes and paired reports, establishing the first benchmarks for report generation from 3D HR medical images. Our method consistently surpasses existing methods on this benchmark across three different settings: normal-resolution, high-resolution inputs, and zero-shot domain transfer, all at an acceptable computational cost, trainable on a single A100-80G.
Authors: Ming Hu, Peng Xia, Lin Wang, Siyuan Yan, Feilong Tang, Zhongxing Xu, Yimin Luo, Kaimin Song, Jurgen Leitner, Xuelian Cheng, Jun Cheng, Chi Liu, Kaijing Zhou, Zongyuan Ge
Abstract: Surgical scene perception via videos are critical for advancing robotic surgery, telesurgery, and AI-assisted surgery, particularly in ophthalmology. However, the scarcity of diverse and richly annotated video datasets has hindered the development of intelligent systems for surgical workflow analysis. Existing datasets for surgical workflow analysis, which typically face challenges such as small scale, a lack of diversity in surgery and phase categories, and the absence of time-localized annotations, limit the requirements for action understanding and model generalization validation in complex and diverse real-world surgical scenarios. To address this gap, we introduce OphNet, a large-scale, expert-annotated video benchmark for ophthalmic surgical workflow understanding. OphNet features: 1) A diverse collection of 2,278 surgical videos spanning 66 types of cataract, glaucoma, and corneal surgeries, with detailed annotations for 102 unique surgical phases and 150 granular operations; 2) It offers sequential and hierarchical annotations for each surgery, phase, and operation, enabling comprehensive understanding and improved interpretability; 3) Moreover, OphNet provides time-localized annotations, facilitating temporal localization and prediction tasks within surgical workflows. With approximately 205 hours of surgical videos, OphNet is about 20 times larger than the largest existing surgical workflow analysis benchmark. Our dataset and code have been made available at: \url{https://github.com/minghu0830/OphNet-benchmark}.
Authors: Md. Shariful Islam, SM Shaqib, Shahriar Sultan Ramit, Shahrun Akter Khushbu, Mr. Abdus Sattar, Dr. Sheak Rashed Haider Noori
Abstract: In the construction sector, ensuring worker safety is of the utmost significance. In this study, a deep learning-based technique is presented for identifying safety gear worn by construction workers, such as helmets, goggles, jackets, gloves, and footwears. The recommended approach uses the YOLO v7 (You Only Look Once) object detection algorithm to precisely locate these safety items. The dataset utilized in this work consists of labeled images split into training, testing and validation sets. Each image has bounding box labels that indicate where the safety equipment is located within the image. The model is trained to identify and categorize the safety equipment based on the labeled dataset through an iterative training approach. We used custom dataset to train this model. Our trained model performed admirably well, with good precision, recall, and F1-score for safety equipment recognition. Also, the model's evaluation produced encouraging results, with a mAP@0.5 score of 87.7\%. The model performs effectively, making it possible to quickly identify safety equipment violations on building sites. A thorough evaluation of the outcomes reveals the model's advantages and points up potential areas for development. By offering an automatic and trustworthy method for safety equipment detection, this research makes a contribution to the fields of computer vision and workplace safety. The proposed deep learning-based approach will increase safety compliance and reduce the risk of accidents in the construction industry
Authors: Lixian Zhang, Yi Zhao, Runmin Dong, Jinxiao Zhang, Shuai Yuan, Shilei Cao, Mengxuan Chen, Juepeng Zheng, Weijia Li, Wei Liu, Litong Feng, Haohuan Fu
Abstract: Vast amounts of remote sensing (RS) data provide Earth observations across multiple dimensions, encompassing critical spatial, temporal, and spectral information which is essential for addressing global-scale challenges such as land use monitoring, disaster prevention, and environmental change mitigation. Despite various pre-training methods tailored to the characteristics of RS data, a key limitation persists: the inability to effectively integrate spatial, temporal, and spectral information within a single unified model. To unlock the potential of RS data, we construct a Spatial-Temporal-Spectral Structured Dataset (STSSD) characterized by the incorporation of multiple RS sources, diverse coverage, unified locations within image sets, and heterogeneity within images. Building upon this structured dataset, we propose an Anchor-Aware Masked AutoEncoder method (A$^{2}$-MAE), leveraging intrinsic complementary information from the different kinds of images and geo-information to reconstruct the masked patches during the pre-training phase. A$^{2}$-MAE integrates an anchor-aware masking strategy and a geographic encoding module to comprehensively exploit the properties of RS images. Specifically, the proposed anchor-aware masking strategy dynamically adapts the masking process based on the meta-information of a pre-selected anchor image, thereby facilitating the training on images captured by diverse types of RS sources within one model. Furthermore, we propose a geographic encoding method to leverage accurate spatial patterns, enhancing the model generalization capabilities for downstream applications that are generally location-related. Extensive experiments demonstrate our method achieves comprehensive improvements across various downstream tasks compared with existing RS pre-training methods, including image classification, semantic segmentation, and change detection tasks.
Authors: Weizhao He, Yang Zhang, Wei Zhuo, Linlin Shen, Jiaqi Yang, Songhe Deng, Liang Sun
Abstract: Few-shot semantic segmentation (FSS) endeavors to segment unseen classes with only a few labeled samples. Current FSS methods are commonly built on the assumption that their training and application scenarios share similar domains, and their performances degrade significantly while applied to a distinct domain. To this end, we propose to leverage the cutting-edge foundation model, the Segment Anything Model (SAM), for generalization enhancement. The SAM however performs unsatisfactorily on domains that are distinct from its training data, which primarily comprise natural scene images, and it does not support automatic segmentation of specific semantics due to its interactive prompting mechanism. In our work, we introduce APSeg, a novel auto-prompt network for cross-domain few-shot semantic segmentation (CD-FSS), which is designed to be auto-prompted for guiding cross-domain segmentation. Specifically, we propose a Dual Prototype Anchor Transformation (DPAT) module that fuses pseudo query prototypes extracted based on cycle-consistency with support prototypes, allowing features to be transformed into a more stable domain-agnostic space. Additionally, a Meta Prompt Generator (MPG) module is introduced to automatically generate prompt embeddings, eliminating the need for manual visual prompts. We build an efficient model which can be applied directly to target domains without fine-tuning. Extensive experiments on four cross-domain datasets show that our model outperforms the state-of-the-art CD-FSS method by 5.24% and 3.10% in average accuracy on 1-shot and 5-shot settings, respectively.
Authors: Xuehai He, Weixi Feng, Kaizhi Zheng, Yujie Lu, Wanrong Zhu, Jiachen Li, Yue Fan, Jianfeng Wang, Linjie Li, Zhengyuan Yang, Kevin Lin, William Yang Wang, Lijuan Wang, Xin Eric Wang
Abstract: Multimodal Language Language Models (MLLMs) demonstrate the emerging abilities of "world models" -- interpreting and reasoning about complex real-world dynamics. To assess these abilities, we posit videos are the ideal medium, as they encapsulate rich representations of real-world dynamics and causalities. To this end, we introduce MMWorld, a new benchmark for multi-discipline, multi-faceted multimodal video understanding. MMWorld distinguishes itself from previous video understanding benchmarks with two unique advantages: (1) multi-discipline, covering various disciplines that often require domain expertise for comprehensive understanding; (2) multi-faceted reasoning, including explanation, counterfactual thinking, future prediction, etc. MMWorld consists of a human-annotated dataset to evaluate MLLMs with questions about the whole videos and a synthetic dataset to analyze MLLMs within a single modality of perception. Together, MMWorld encompasses 1,910 videos across seven broad disciplines and 69 subdisciplines, complete with 6,627 question-answer pairs and associated captions. The evaluation includes 2 proprietary and 10 open-source MLLMs, which struggle on MMWorld (e.g., GPT-4V performs the best with only 52.3\% accuracy), showing large room for improvement. Further ablation studies reveal other interesting findings such as models' different skill sets from humans. We hope MMWorld can serve as an essential step towards world model evaluation in videos.
Authors: Zhao Mandi, Yijia Weng, Dominik Bauer, Shuran Song
Abstract: We present Real2Code, a novel approach to reconstructing articulated objects via code generation. Given visual observations of an object, we first reconstruct its part geometry using an image segmentation model and a shape completion model. We then represent the object parts with oriented bounding boxes, which are input to a fine-tuned large language model (LLM) to predict joint articulation as code. By leveraging pre-trained vision and language models, our approach scales elegantly with the number of articulated parts, and generalizes from synthetic training data to real world objects in unstructured environments. Experimental results demonstrate that Real2Code significantly outperforms previous state-of-the-art in reconstruction accuracy, and is the first approach to extrapolate beyond objects' structural complexity in the training set, and reconstructs objects with up to 10 articulated parts. When incorporated with a stereo reconstruction model, Real2Code also generalizes to real world objects from a handful of multi-view RGB images, without the need for depth or camera information.
Authors: Yi-Fan Zhang, Qingsong Wen, Chaoyou Fu, Xue Wang, Zhang Zhang, Liang Wang, Rong Jin
Abstract: Seeing clearly with high resolution is a foundation of Large Multimodal Models (LMMs), which has been proven to be vital for visual perception and reasoning. Existing works usually employ a straightforward resolution upscaling method, where the image consists of global and local branches, with the latter being the sliced image patches but resized to the same resolution as the former. This means that higher resolution requires more local patches, resulting in exorbitant computational expenses, and meanwhile, the dominance of local image tokens may diminish the global context. In this paper, we dive into the problems and propose a new framework as well as an elaborate optimization strategy. Specifically, we extract contextual information from the global view using a mixture of adapters, based on the observation that different adapters excel at different tasks. With regard to local patches, learnable query embeddings are introduced to reduce image tokens, the most important tokens accounting for the user question will be further selected by a similarity-based selector. Our empirical results demonstrate a `less is more' pattern, where \textit{utilizing fewer but more informative local image tokens leads to improved performance}. Besides, a significant challenge lies in the training strategy, as simultaneous end-to-end training of the global mining block and local compression block does not yield optimal results. We thus advocate for an alternating training way, ensuring balanced learning between global and local aspects. Finally, we also introduce a challenging dataset with high requirements for image detail, enhancing the training of the local compression layer. The proposed method, termed LMM with Sophisticated Tasks, Local image compression, and Mixture of global Experts (SliME), achieves leading performance across various benchmarks with only 2 million training data.
Authors: Paul Pu Liang, Chun Kai Ling, Yun Cheng, Alex Obolenskiy, Yudong Liu, Rohan Pandey, Alex Wilf, Louis-Philippe Morency, Ruslan Salakhutdinov
Abstract: In many machine learning systems that jointly learn from multiple modalities, a core research question is to understand the nature of multimodal interactions: how modalities combine to provide new task-relevant information that was not present in either alone. We study this challenge of interaction quantification in a semi-supervised setting with only labeled unimodal data and naturally co-occurring multimodal data (e.g., unlabeled images and captions, video and corresponding audio) but when labeling them is time-consuming. Using a precise information-theoretic definition of interactions, our key contribution is the derivation of lower and upper bounds to quantify the amount of multimodal interactions in this semi-supervised setting. We propose two lower bounds: one based on the shared information between modalities and the other based on disagreement between separately trained unimodal classifiers, and derive an upper bound through connections to approximate algorithms for min-entropy couplings. We validate these estimated bounds and show how they accurately track true interactions. Finally, we show how these theoretical results can be used to estimate multimodal model performance, guide data collection, and select appropriate multimodal models for various tasks.
Authors: Wentao Liu, Tong Tian, Lemeng Wang, Weijin Xu, Lei Li, Haoyuan Li, Wenyi Zhao, Siyu Tian, Xipeng Pan, Huihua Yang, Feng Gao, Yiming Deng, Xin Yang, Ruisheng Su
Abstract: The automated segmentation of Intracranial Arteries (IA) in Digital Subtraction Angiography (DSA) plays a crucial role in the quantification of vascular morphology, significantly contributing to computer-assisted stroke research and clinical practice. Current research primarily focuses on the segmentation of single-frame DSA using proprietary datasets. However, these methods face challenges due to the inherent limitation of single-frame DSA, which only partially displays vascular contrast, thereby hindering accurate vascular structure representation. In this work, we introduce DIAS, a dataset specifically developed for IA segmentation in DSA sequences. We establish a comprehensive benchmark for evaluating DIAS, covering full, weak, and semi-supervised segmentation methods. Specifically, we propose the vessel sequence segmentation network, in which the sequence feature extraction module effectively captures spatiotemporal representations of intravascular contrast, achieving intracranial artery segmentation in 2D+Time DSA sequences. For weakly-supervised IA segmentation, we propose a novel scribble learning-based image segmentation framework, which, under the guidance of scribble labels, employs cross pseudo-supervision and consistency regularization to improve the performance of the segmentation network. Furthermore, we introduce the random patch-based self-training framework, aimed at alleviating the performance constraints encountered in IA segmentation due to the limited availability of annotated DSA data. Our extensive experiments on the DIAS dataset demonstrate the effectiveness of these methods as potential baselines for future research and clinical applications. The dataset and code are publicly available at https://doi.org/10.5281/zenodo.11396520 and https://github.com/lseventeen/DIAS.
URLs: https://doi.org/10.5281/zenodo.11396520, https://github.com/lseventeen/DIAS.
Authors: Anith Selvakumar, Homa Fashandi
Abstract: Distance Metric Learning (DML) has typically dominated the audio-visual speaker verification problem space, owing to strong performance in new and unseen classes. In our work, we explored multitask learning techniques to further enhance DML, and show that an auxiliary task with even weak labels can increase the quality of the learned speaker representation without increasing model complexity during inference. We also extend the Generalized End-to-End Loss (GE2E) to multimodal inputs and demonstrate that it can achieve competitive performance in an audio-visual space. Finally, we introduce AV-Mixup, a multimodal augmentation technique during training time that has shown to reduce speaker overfit. Our network achieves state of the art performance for speaker verification, reporting 0.244%, 0.252%, 0.441% Equal Error Rate (EER) on the VoxCeleb1-O/E/H test sets, which is to our knowledge, the best published results on VoxCeleb1-E and VoxCeleb1-H.
Authors: Shuai Wang, Eric Nalisnick
Abstract: We apply active learning to help with data scarcity problems in sign languages. In particular, we perform a novel analysis of the effect of pre-training. Since many sign languages are linguistic descendants of French sign language, they share hand configurations, which pre-training can hopefully exploit. We test this hypothesis on American, Chinese, German, and Irish fingerspelling corpora. We do observe a benefit from pre-training, but this may be due to visual rather than linguistic similarities
Authors: Xiang Yue, Yuansheng Ni, Kai Zhang, Tianyu Zheng, Ruoqi Liu, Ge Zhang, Samuel Stevens, Dongfu Jiang, Weiming Ren, Yuxuan Sun, Cong Wei, Botao Yu, Ruibin Yuan, Renliang Sun, Ming Yin, Boyuan Zheng, Zhenzhu Yang, Yibo Liu, Wenhao Huang, Huan Sun, Yu Su, Wenhu Chen
Abstract: We introduce MMMU: a new benchmark designed to evaluate multimodal models on massive multi-discipline tasks demanding college-level subject knowledge and deliberate reasoning. MMMU includes 11.5K meticulously collected multimodal questions from college exams, quizzes, and textbooks, covering six core disciplines: Art & Design, Business, Science, Health & Medicine, Humanities & Social Science, and Tech & Engineering. These questions span 30 subjects and 183 subfields, comprising 30 highly heterogeneous image types, such as charts, diagrams, maps, tables, music sheets, and chemical structures. Unlike existing benchmarks, MMMU focuses on advanced perception and reasoning with domain-specific knowledge, challenging models to perform tasks akin to those faced by experts. The evaluation of 14 open-source LMMs as well as the proprietary GPT-4V(ision) and Gemini highlights the substantial challenges posed by MMMU. Even the advanced GPT-4V and Gemini Ultra only achieve accuracies of 56% and 59% respectively, indicating significant room for improvement. We believe MMMU will stimulate the community to build next-generation multimodal foundation models towards expert artificial general intelligence.
Authors: Riccardo Zaccone, Carlo Masone, Marco Ciccone
Abstract: Federated Learning (FL) has emerged as the state-of-the-art approach for learning from decentralized data in privacy-constrained scenarios. However, system and statistical challenges hinder real-world applications, which demand efficient learning from edge devices and robustness to heterogeneity. Despite significant research efforts, existing approaches (i) are not sufficiently robust, (ii) do not perform well in large-scale scenarios, and (iii) are not communication efficient. In this work, we propose a novel Generalized Heavy-Ball Momentum (GHBM), motivating its principled application to counteract the effects of statistical heterogeneity in FL. Then, we present FedHBM as an adaptive, communication-efficient by-design instance of GHBM. Extensive experimentation on vision and language tasks, in both controlled and realistic large-scale scenarios, provides compelling evidence of substantial and consistent performance gains over the state of the art.
Authors: Han Huang, Haitian Zhong, Tao Yu, Qiang Liu, Shu Wu, Liang Wang, Tieniu Tan
Abstract: Recently, knowledge editing on large language models (LLMs) has received considerable attention. Compared to this, editing Large Vision-Language Models (LVLMs) faces extra challenges from diverse data modalities and complicated model components, and data for LVLMs editing are limited. The existing LVLM editing benchmark, which comprises three metrics (Reliability, Locality, and Generality), falls short in the quality of synthesized evaluation images and cannot assess whether models apply edited knowledge in relevant content. Therefore, we employ more reliable data collection methods to construct a new Large $\textbf{V}$ision-$\textbf{L}$anguage Model $\textbf{K}$nowledge $\textbf{E}$diting $\textbf{B}$enchmark, $\textbf{VLKEB}$, and extend the Portability metric for more comprehensive evaluation. Leveraging a multi-modal knowledge graph, our image data are bound with knowledge entities. This can be further used to extract entity-related knowledge, which constitutes the base of editing data. We conduct experiments of different editing methods on five LVLMs, and thoroughly analyze how do they impact the models. The results reveal strengths and deficiencies of these methods and hopefully provide insights for future research. The codes and dataset are available at: $\href{https://github.com/VLKEB/VLKEB}{\text{https://github.com/VLKEB/VLKEB}}$.
URLs: https://github.com/VLKEB/VLKEB, https://github.com/VLKEB/VLKEB
Authors: Zhiyuan Li, Yanhui Zhou, Hao Wei, Chenyang Ge, Jingwen Jiang
Abstract: Image compression at extremely low bitrates (below 0.1 bits per pixel (bpp)) is a significant challenge due to substantial information loss. In this work, we propose a novel two-stage extreme image compression framework that exploits the powerful generative capability of pre-trained diffusion models to achieve realistic image reconstruction at extremely low bitrates. In the first stage, we treat the latent representation of images in the diffusion space as guidance, employing a VAE-based compression approach to compress images and initially decode the compressed information into content variables. The second stage leverages pre-trained stable diffusion to reconstruct images under the guidance of content variables. Specifically, we introduce a small control module to inject content information while keeping the stable diffusion model fixed to maintain its generative capability. Furthermore, we design a space alignment loss to force the content variables to align with the diffusion space and provide the necessary constraints for optimization. Extensive experiments demonstrate that our method significantly outperforms state-of-the-art approaches in terms of visual performance at extremely low bitrates.
Authors: Xiaofeng Zhang, Chen Shen, Xiaosong Yuan, Shaotian Yan, Liang Xie, Wenxiao Wang, Chaochen Gu, Hao Tang, Jieping Ye
Abstract: Recently, multimodal large language models have exploded with an endless variety, most of the popular Large Vision Language Models (LVLMs) depend on sequential visual representation, where images are converted into hundreds or thousands of tokens before being input into the Large Language Model (LLM) along with language prompts. The black-box design hinders the interpretability of visual-language models, especially regarding more complex reasoning tasks. To explore the interaction process between image and text in complex reasoning tasks, we introduce the information flow method to visualize the interaction mechanism. By analyzing the dynamic flow of the information flow, we find that the information flow appears to converge in the shallow layer. Further investigation revealed a redundancy of the image token in the shallow layer. Consequently, a truncation strategy was introduced to aggregate image tokens within these shallow layers. This approach has been validated through experiments across multiple models, yielding consistent improvements.
Authors: Michel Adamic, Wellington Avelino, Anna Brandenberger, Bryan Chiang, Hunter Davis, Stephen Fay, Andrew Gregory, Aayush Gupta, Raphael Hotter, Grace Jiang, Fiona Leng, Stephen Polcyn, Thomas Ribeiro, Paul Scotti, Michelle Wang, Marley Xiong, Jonathan Xu
Abstract: We demonstrate the possibility of reconstructing images from fNIRS brain activity and start building a prototype to match the required specs. By training an image reconstruction model on downsampled fMRI data, we discovered that cm-scale spatial resolution is sufficient for image generation. We obtained 71% retrieval accuracy with 1-cm resolution, compared to 93% on the full-resolution fMRI, and 20% with 2-cm resolution. With simulations and high-density tomography, we found that time-domain fNIRS can achieve 1-cm resolution, compared to 2-cm resolution for continuous-wave fNIRS. Lastly, we share designs for a prototype time-domain fNIRS device, consisting of a laser driver, a single photon detector, and a time-to-digital converter system.