Authors: Sheng Cheng, Tejas Gokhale, Yezhou Yang
Generalizing to unseen image domains is a challenging problem primarily due to the lack of diverse training data, inaccessible target data, and the large domain shift that may exist in many real-world settings. As such data augmentation is a critical component of domain generalization methods that seek to address this problem. We present Adversarial Bayesian Augmentation (ABA), a novel algorithm that learns to generate image augmentations in the challenging single-source domain generalization setting. ABA draws on the strengths of adversarial learning and Bayesian neural networks to guide the generation of diverse data augmentations -- these synthesized image domains aid the classifier in generalizing to unseen domains. We demonstrate the strength of ABA on several types of domain shift including style shift, subpopulation shift, and shift in the medical imaging setting. ABA outperforms all previous state-of-the-art methods, including pre-specified augmentations, pixel-based and convolutional-based augmentations.
Authors: Pratyush Maini, Michael C. Mozer, Hanie Sedghi, Zachary C. Lipton, J. Zico Kolter, Chiyuan Zhang
Recent efforts at explaining the interplay of memorization and generalization in deep overparametrized networks have posited that neural networks $\textit{memorize}$ "hard" examples in the final few layers of the model. Memorization refers to the ability to correctly predict on $\textit{atypical}$ examples of the training set. In this work, we show that rather than being confined to individual layers, memorization is a phenomenon confined to a small set of neurons in various layers of the model. First, via three experimental sources of converging evidence, we find that most layers are redundant for the memorization of examples and the layers that contribute to example memorization are, in general, not the final layers. The three sources are $\textit{gradient accounting}$ (measuring the contribution to the gradient norms from memorized and clean examples), $\textit{layer rewinding}$ (replacing specific model weights of a converged model with previous training checkpoints), and $\textit{retraining}$ (training rewound layers only on clean examples). Second, we ask a more generic question: can memorization be localized $\textit{anywhere}$ in a model? We discover that memorization is often confined to a small number of neurons or channels (around 5) of the model. Based on these insights we propose a new form of dropout -- $\textit{example-tied dropout}$ that enables us to direct the memorization of examples to an apriori determined set of neurons. By dropping out these neurons, we are able to reduce the accuracy on memorized examples from $100\%\to3\%$, while also reducing the generalization gap.
Authors: Saurav Sharma, Chinedu Innocent Nwoye, Didier Mutter, Nicolas Padoy
Surgical action triplets describe instrument-tissue interactions as (instrument, verb, target) combinations, thereby supporting a detailed analysis of surgical scene activities and workflow. This work focuses on surgical action triplet detection, which is challenging but more precise than the traditional triplet recognition task as it consists of joint (1) localization of surgical instruments and (2) recognition of the surgical action triplet associated with every localized instrument. Triplet detection is highly complex due to the lack of spatial triplet annotation. We analyze how the amount of instrument spatial annotations affects triplet detection and observe that accurate instrument localization does not guarantee better triplet detection due to the risk of erroneous associations with the verbs and targets. To solve the two tasks, we propose MCIT-IG, a two-stage network, that stands for Multi-Class Instrument-aware Transformer-Interaction Graph. The MCIT stage of our network models per class embedding of the targets as additional features to reduce the risk of misassociating triplets. Furthermore, the IG stage constructs a bipartite dynamic graph to model the interaction between the instruments and targets, cast as the verbs. We utilize a mixed-supervised learning strategy that combines weak target presence labels for MCIT and pseudo triplet labels for IG to train our network. We observed that complementing minimal instrument spatial annotations with target embeddings results in better triplet detection. We evaluate our model on the CholecT50 dataset and show improved performance on both instrument localization and triplet detection, topping the leaderboard of the CholecTriplet challenge in MICCAI 2022.
Authors: Anagh Malik, Parsa Mirdehghan, Sotiris Nousias, Kiriakos N. Kutulakos, David B. Lindell
Neural radiance fields (NeRFs) have become a ubiquitous tool for modeling scene appearance and geometry from multiview imagery. Recent work has also begun to explore how to use additional supervision from lidar or depth sensor measurements in the NeRF framework. However, previous lidar-supervised NeRFs focus on rendering conventional camera imagery and use lidar-derived point cloud data as auxiliary supervision; thus, they fail to incorporate the underlying image formation model of the lidar. Here, we propose a novel method for rendering transient NeRFs that take as input the raw, time-resolved photon count histograms measured by a single-photon lidar system, and we seek to render such histograms from novel views. Different from conventional NeRFs, the approach relies on a time-resolved version of the volume rendering equation to render the lidar measurements and capture transient light transport phenomena at picosecond timescales. We evaluate our method on a first-of-its-kind dataset of simulated and captured transient multiview scans from a prototype single-photon lidar. Overall, our work brings NeRFs to a new dimension of imaging at transient timescales, newly enabling rendering of transient imagery from novel views. Additionally, we show that our approach recovers improved geometry and conventional appearance compared to point cloud-based supervision when training on few input viewpoints. Transient NeRFs may be especially useful for applications which seek to simulate raw lidar measurements for downstream tasks in autonomous driving, robotics, and remote sensing.
Authors: Pierre Le Jeune, Anissa Mokraoui
In Few-Shot Object Detection (FSOD), detecting small objects is extremely difficult. The limited supervision cripples the localization capabilities of the models and a few pixels shift can dramatically reduce the Intersection over Union (IoU) between the ground truth and predicted boxes for small objects. To this end, we propose Scale-adaptive Intersection over Union (SIoU), a novel box similarity measure. SIoU changes with the objects' size, it is more lenient with small object shifts. We conducted a user study and SIoU better aligns than IoU with human judgment. Employing SIoU as an evaluation criterion helps to build more user-oriented models. SIoU can also be used as a loss function to prioritize small objects during training, outperforming existing loss functions. SIoU improves small object detection in the non-few-shot regime, but this setting is unrealistic in the industry as annotated detection datasets are often too expensive to acquire. Hence, our experiments mainly focus on the few-shot regime to demonstrate the superiority and versatility of SIoU loss. SIoU improves significantly FSOD performance on small objects in both natural (Pascal VOC and COCO datasets) and aerial images (DOTA and DIOR). In aerial imagery, small objects are critical and SIoU loss achieves new state-of-the-art FSOD on DOTA and DIOR.
Authors: Jingwei Zhang, Ke Ma, Saarthak Kapse, Joel Saltz, Maria Vakalopoulou, Prateek Prasanna, Dimitris Samaras
Semantic segmentations of pathological entities have crucial clinical value in computational pathology workflows. Foundation models, such as the Segment Anything Model (SAM), have been recently proposed for universal use in segmentation tasks. SAM shows remarkable promise in instance segmentation on natural images. However, the applicability of SAM to computational pathology tasks is limited due to the following factors: (1) lack of comprehensive pathology datasets used in SAM training and (2) the design of SAM is not inherently optimized for semantic segmentation tasks. In this work, we adapt SAM for semantic segmentation by introducing trainable class prompts, followed by further enhancements through the incorporation of a pathology encoder, specifically a pathology foundation model. Our framework, SAM-Path enhances SAM's ability to conduct semantic segmentation in digital pathology without human input prompts. Through experiments on two public pathology datasets, the BCSS and the CRAG datasets, we demonstrate that the fine-tuning with trainable class prompts outperforms vanilla SAM with manual prompts and post-processing by 27.52% in Dice score and 71.63% in IOU. On these two datasets, the proposed additional pathology foundation model further achieves a relative improvement of 5.07% to 5.12% in Dice score and 4.50% to 8.48% in IOU.
Authors: Shuangbing Song, Fan Zhong, Tianju Wang, Xueying Qin, Changhe Tu
Guided upsampling is an effective approach for accelerating high-resolution image processing. In this paper, we propose a simple yet effective guided upsampling method. Each pixel in the high-resolution image is represented as a linear interpolation of two low-resolution pixels, whose indices and weights are optimized to minimize the upsampling error. The downsampling can be jointly optimized in order to prevent missing small isolated regions. Our method can be derived from the color line model and local color transformations. Compared to previous methods, our method can better preserve detail effects while suppressing artifacts such as bleeding and blurring. It is efficient, easy to implement, and free of sensitive parameters. We evaluate the proposed method with a wide range of image operators, and show its advantages through quantitative and qualitative analysis. We demonstrate the advantages of our method for both interactive image editing and real-time high-resolution video processing. In particular, for interactive editing, the joint optimization can be precomputed, thus allowing for instant feedback without hardware acceleration.
Authors: Lars Nieradzik, Jördis Sieburg-Rockel, Stephanie Helmling, Janis Keuper, Thomas Weibel, Andrea Olbrich, Henrike Stephani
We have developed a methodology for the systematic generation of a large image dataset of macerated wood references, which we used to generate image data for nine hardwood genera. This is the basis for a substantial approach to automate, for the first time, the identification of hardwood species in microscopic images of fibrous materials by deep learning. Our methodology includes a flexible pipeline for easy annotation of vessel elements. We compare the performance of different neural network architectures and hyperparameters. Our proposed method performs similarly well to human experts. In the future, this will improve controls on global wood fiber product flows to protect forests.
Authors: Sabine Muzellec, Leo Andeol, Thomas Fel, Rufin VanRullen, Thomas Serre
Recent years have witnessed an explosion in the development of novel prediction-based attribution methods, which have slowly been supplanting older gradient-based methods to explain the decisions of deep neural networks. However, it is still not clear why prediction-based methods outperform gradient-based ones. Here, we start with an empirical observation: these two approaches yield attribution maps with very different power spectra, with gradient-based methods revealing more high-frequency content than prediction-based methods. This observation raises multiple questions: What is the source of this high-frequency information, and does it truly reflect decisions made by the system? Lastly, why would the absence of high-frequency information in prediction-based methods yield better explainability scores along multiple metrics? We analyze the gradient of three representative visual classification models and observe that it contains noisy information emanating from high-frequencies. Furthermore, our analysis reveals that the operations used in Convolutional Neural Networks (CNNs) for downsampling appear to be a significant source of this high-frequency content -- suggesting aliasing as a possible underlying basis. We then apply an optimal low-pass filter for attribution maps and demonstrate that it improves gradient-based attribution methods. We show that (i) removing high-frequency noise yields significant improvements in the explainability scores obtained with gradient-based methods across multiple models -- leading to (ii) a novel ranking of state-of-the-art methods with gradient-based methods at the top. We believe that our results will spur renewed interest in simpler and computationally more efficient gradient-based methods for explainability.
Authors: Zhaoxin Fan, Puquan Pan, Zeren Zhang, Ce Chen, Tianyang Wang, Siyang Zheng, Min Xu
Few-shot medical image semantic segmentation is of paramount importance in the domain of medical image analysis. However, existing methodologies grapple with the challenge of data scarcity during the training phase, leading to over-fitting. To mitigate this issue, we introduce a novel Unsupervised Dense Few-shot Medical Image Segmentation Model Training Pipeline (DenseMP) that capitalizes on unsupervised dense pre-training. DenseMP is composed of two distinct stages: (1) segmentation-aware dense contrastive pre-training, and (2) few-shot-aware superpixel guided dense pre-training. These stages collaboratively yield a pre-trained initial model specifically designed for few-shot medical image segmentation, which can subsequently be fine-tuned on the target dataset. Our proposed pipeline significantly enhances the performance of the widely recognized few-shot segmentation model, PA-Net, achieving state-of-the-art results on the Abd-CT and Abd-MRI datasets. Code will be released after acceptance.
Authors: Md. Mahfuzur Rahman, Vince D. Calhoun, Sergey M. Plis
Deep learning (DL) models have been popular due to their ability to learn directly from the raw data in an end-to-end paradigm, alleviating the concern of a separate error-prone feature extraction phase. Recent DL-based neuroimaging studies have also witnessed a noticeable performance advancement over traditional machine learning algorithms. But the challenges of deep learning models still exist because of the lack of transparency in these models for their successful deployment in real-world applications. In recent years, Explainable AI (XAI) has undergone a surge of developments mainly to get intuitions of how the models reached the decisions, which is essential for safety-critical domains such as healthcare, finance, and law enforcement agencies. While the interpretability domain is advancing noticeably, researchers are still unclear about what aspect of model learning a post hoc method reveals and how to validate its reliability. This paper comprehensively reviews interpretable deep learning models in the neuroimaging domain. Firstly, we summarize the current status of interpretability resources in general, focusing on the progression of methods, associated challenges, and opinions. Secondly, we discuss how multiple recent neuroimaging studies leveraged model interpretability to capture anatomical and functional brain alterations most relevant to model predictions. Finally, we discuss the limitations of the current practices and offer some valuable insights and guidance on how we can steer our future research directions to make deep learning models substantially interpretable and thus advance scientific understanding of brain disorders.
Authors: Ka Chun Shum, Hong-Wing Pang, Binh-Son Hua, Duc Thanh Nguyen, Sai-Kit Yeung
In this paper, we address the problem of conditional scene decoration for 360-degree images. Our method takes a 360-degree background photograph of an indoor scene and generates decorated images of the same scene in the panorama view. To do this, we develop a 360-aware object layout generator that learns latent object vectors in the 360-degree view to enable a variety of furniture arrangements for an input 360-degree background image. We use this object layout to condition a generative adversarial network to synthesize images of an input scene. To further reinforce the generation capability of our model, we develop a simple yet effective scene emptier that removes the generated furniture and produces an emptied scene for our model to learn a cyclic constraint. We train the model on the Structure3D dataset and show that our model can generate diverse decorations with controllable object layout. Our method achieves state-of-the-art performance on the Structure3D dataset and generalizes well to the Zillow indoor scene dataset. Our user study confirms the immersive experiences provided by the realistic image quality and furniture layout in our generation results. Our implementation will be made available.
Authors: Ehsan Qasemi, Jonathan M. Francis, Alessandro Oltramari
Video Question Answering (VidQA) exhibits remarkable potential in facilitating advanced machine reasoning capabilities within the domains of Intelligent Traffic Monitoring and Intelligent Transportation Systems. Nevertheless, the integration of urban traffic scene knowledge into VidQA systems has received limited attention in previous research endeavors. In this work, we present a novel approach termed Traffic-domain Video Question Answering with Automatic Captioning (TRIVIA), which serves as a weak-supervision technique for infusing traffic-domain knowledge into large video-language models. Empirical findings obtained from the SUTD-TrafficQA task highlight the substantial enhancements achieved by TRIVIA, elevating the accuracy of representative video-language models by a remarkable 6.5 points (19.88%) compared to baseline settings. This pioneering methodology holds great promise for driving advancements in the field, inspiring researchers and practitioners alike to unlock the full potential of emerging video-language models in traffic-related applications.
Authors: Wei-Lun Huang, Davood Tashayyod, Jun Kang, Amir Gandjbakhche, Michael Kazhdan, Mehran Armand
Longitudinal tracking of skin lesions - finding correspondence, changes in morphology, and texture - is beneficial to the early detection of melanoma. However, it has not been well investigated in the context of full-body imaging. We propose a novel framework combining geometric and texture information to localize skin lesion correspondence from a source scan to a target scan in total body photography (TBP). Body landmarks or sparse correspondence are first created on the source and target 3D textured meshes. Every vertex on each of the meshes is then mapped to a feature vector characterizing the geodesic distances to the landmarks on that mesh. Then, for each lesion of interest (LOI) on the source, its corresponding location on the target is first coarsely estimated using the geometric information encoded in the feature vectors and then refined using the texture information. We evaluated the framework quantitatively on both a public and a private dataset, for which our success rates (at 10 mm criterion) are comparable to the only reported longitudinal study. As full-body 3D capture becomes more prevalent and has higher quality, we expect the proposed method to constitute a valuable step in the longitudinal tracking of skin lesions.
Authors: Francesco Tonini, Nicola Dall'Asen, Cigdem Beyan, Elisa Ricci
Gaze target detection aims to predict the image location where the person is looking and the probability that a gaze is out of the scene. Several works have tackled this task by regressing a gaze heatmap centered on the gaze location, however, they overlooked decoding the relationship between the people and the gazed objects. This paper proposes a Transformer-based architecture that automatically detects objects (including heads) in the scene to build associations between every head and the gazed-head/object, resulting in a comprehensive, explainable gaze analysis composed of: gaze target area, gaze pixel point, the class and the image location of the gazed-object. Upon evaluation of the in-the-wild benchmarks, our method achieves state-of-the-art results on all metrics (up to 2.91% gain in AUC, 50% reduction in gaze distance, and 9% gain in out-of-frame average precision) for gaze target detection and 11-13% improvement in average precision for the classification and the localization of the gazed-objects. The code of the proposed method is available https://github.com/francescotonini/object-aware-gaze-target-detection
Authors: Jinlong Li, Runsheng Xu, Jin Ma, Qin Zou, Jiaqi Ma, Hongkai Yu
Most object detection models for autonomous driving may experience a significant drop in performance when deployed in real-world applications, due to the well-known domain shift issue. Supervised object detection methods for autonomous driving usually assume a consistent feature distribution between training and testing data, however, such assumptions may not always be the case when weather conditions differ significantly. For example, an object detection model trained under clear weather may not perform well in foggy or rainy weather, due to the domain gap. Overcoming detection bottlenecks in foggy or rainy weather scenarios is a significant challenge for autonomous vehicles deployed in the wild. To address the domain gap in different weather conditions, This paper proposes a novel domain adaptive object detection framework for autonomous driving in foggy and rainy weather. Our method leverages both image-level and object-level adaptation to reduce the domain discrepancy in image style and object appearance. Additionally, to enhance the model's performance under challenging samples, we introduce a new adversarial gradient reversal layer that performs adversarial mining on hard examples alongside domain adaptation. Moreover, we propose to generate an auxiliary domain by data augmentation to enforce a new domain-level metric regularization. Experimental results on public benchmarks demonstrate that object detection performance is significantly improved when using our proposed method in domain shift scenarios for autonomous driving applications.
Authors: Liu He, Daniel Aliaga
Modeling and designing urban building layouts is of significant interest in computer vision, computer graphics, and urban applications. A building layout consists of a set of buildings in city blocks defined by a network of roads. We observe that building layouts are discrete structures, consisting of multiple rows of buildings of various shapes, and are amenable to skeletonization for mapping arbitrary city block shapes to a canonical form. Hence, we propose a fully automatic approach to building layout generation using graph attention networks. Our method generates realistic urban layouts given arbitrary road networks, and enables conditional generation based on learned priors. Our results, including user study, demonstrate superior performance as compared to prior layout generation networks, support arbitrary city block and varying building shapes as demonstrated by generating layouts for 28 large cities.
Authors: Bin Duan, Ming Zhong, Yan Yan
With recent advances in computing hardware and surges of deep-learning architectures, learning-based deep image registration methods have surpassed their traditional counterparts, in terms of metric performance and inference time. However, these methods focus on improving performance measurements such as Dice, resulting in less attention given to model behaviors that are equally desirable for registrations, especially for medical imaging. This paper investigates these behaviors for popular learning-based deep registrations under a sanity-checking microscope. We find that most existing registrations suffer from low inverse consistency and nondiscrimination of identical pairs due to overly optimized image similarities. To rectify these behaviors, we propose a novel regularization-based sanity-enforcer method that imposes two sanity checks on the deep model to reduce its inverse consistency errors and increase its discriminative power simultaneously. Moreover, we derive a set of theoretical guarantees for our sanity-checked image registration method, with experimental results supporting our theoretical findings and their effectiveness in increasing the sanity of models without sacrificing any performance. Our code and models are available at \url{https://github.com/tuffr5/Saner-deep-registration}.
Authors: Leilei Ma, Dengdi Sun, Lei Wang, Haifang Zhao, Bin Luo
Extracting image semantics effectively and assigning corresponding labels to multiple objects or attributes for natural images is challenging due to the complex scene contents and confusing label dependencies. Recent works have focused on modeling label relationships with graph and understanding object regions using class activation maps (CAM). However, these methods ignore the complex intra- and inter-category relationships among specific semantic features, and CAM is prone to generate noisy information. To this end, we propose a novel semantic-aware dual contrastive learning framework that incorporates sample-to-sample contrastive learning (SSCL) as well as prototype-to-sample contrastive learning (PSCL). Specifically, we leverage semantic-aware representation learning to extract category-related local discriminative features and construct category prototypes. Then based on SSCL, label-level visual representations of the same category are aggregated together, and features belonging to distinct categories are separated. Meanwhile, we construct a novel PSCL module to narrow the distance between positive samples and category prototypes and push negative samples away from the corresponding category prototypes. Finally, the discriminative label-level features related to the image content are accurately captured by the joint training of the above three parts. Experiments on five challenging large-scale public datasets demonstrate that our proposed method is effective and outperforms the state-of-the-art methods. Code and supplementary materials are released on https://github.com/yu-gi-oh-leilei/SADCL.
Authors: Pengfei Luo, Tong Xu, Shiwei Wu, Chen Zhu, Linli Xu, Enhong Chen
Multimodal entity linking (MEL) task, which aims at resolving ambiguous mentions to a multimodal knowledge graph, has attracted wide attention in recent years. Though large efforts have been made to explore the complementary effect among multiple modalities, however, they may fail to fully absorb the comprehensive expression of abbreviated textual context and implicit visual indication. Even worse, the inevitable noisy data may cause inconsistency of different modalities during the learning process, which severely degenerates the performance. To address the above issues, in this paper, we propose a novel Multi-GraIned Multimodal InteraCtion Network $\textbf{(MIMIC)}$ framework for solving the MEL task. Specifically, the unified inputs of mentions and entities are first encoded by textual/visual encoders separately, to extract global descriptive features and local detailed features. Then, to derive the similarity matching score for each mention-entity pair, we device three interaction units to comprehensively explore the intra-modal interaction and inter-modal fusion among features of entities and mentions. In particular, three modules, namely the Text-based Global-Local interaction Unit (TGLU), Vision-based DuaL interaction Unit (VDLU) and Cross-Modal Fusion-based interaction Unit (CMFU) are designed to capture and integrate the fine-grained representation lying in abbreviated text and implicit visual cues. Afterwards, we introduce a unit-consistency objective function via contrastive learning to avoid inconsistency and model degradation. Experimental results on three public benchmark datasets demonstrate that our solution outperforms various state-of-the-art baselines, and ablation studies verify the effectiveness of designed modules.
Authors: Kibeom Hong, Seogkyu Jeon, Junsoo Lee, Namhyuk Ahn, Kunhee Kim, Pilhyeon Lee, Daesik Kim, Youngjung Uh, Hyeran Byun
To deliver the artistic expression of the target style, recent studies exploit the attention mechanism owing to its ability to map the local patches of the style image to the corresponding patches of the content image. However, because of the low semantic correspondence between arbitrary content and artworks, the attention module repeatedly abuses specific local patches from the style image, resulting in disharmonious and evident repetitive artifacts. To overcome this limitation and accomplish impeccable artistic style transfer, we focus on enhancing the attention mechanism and capturing the rhythm of patterns that organize the style. In this paper, we introduce a novel metric, namely pattern repeatability, that quantifies the repetition of patterns in the style image. Based on the pattern repeatability, we propose Aesthetic Pattern-Aware style transfer Networks (AesPA-Net) that discover the sweet spot of local and global style expressions. In addition, we propose a novel self-supervisory task to encourage the attention mechanism to learn precise and meaningful semantic correspondence. Lastly, we introduce the patch-wise style loss to transfer the elaborate rhythm of local patterns. Through qualitative and quantitative evaluations, we verify the reliability of the proposed pattern repeatability that aligns with human perception, and demonstrate the superiority of the proposed framework.
Authors: Zi Li, Lin Tian, Tony C. W. Mok, Xiaoyu Bai, Puyang Wang, Jia Ge, Jingren Zhou, Le Lu, Xianghua Ye, Ke Yan, Dakai Jin
Estimating displacement vector field via a cost volume computed in the feature space has shown great success in image registration, but it suffers excessive computation burdens. Moreover, existing feature descriptors only extract local features incapable of representing the global semantic information, which is especially important for solving large transformations. To address the discussed issues, we propose SAMConvex, a fast coarse-to-fine discrete optimization method for CT registration that includes a decoupled convex optimization procedure to obtain deformation fields based on a self-supervised anatomical embedding (SAM) feature extractor that captures both local and global information. To be specific, SAMConvex extracts per-voxel features and builds 6D correlation volumes based on SAM features, and iteratively updates a flow field by performing lookups on the correlation volumes with a coarse-to-fine scheme. SAMConvex outperforms the state-of-the-art learning-based methods and optimization-based methods over two inter-patient registration datasets (Abdomen CT and HeadNeck CT) and one intra-patient registration dataset (Lung CT). Moreover, as an optimization-based method, SAMConvex only takes $\sim2$s ($\sim5s$ with instance optimization) for one paired images.
Authors: Ming Tong, Xuefeng Yan, Yongzhen Wang
Visual-based measurement systems are frequently affected by rainy weather due to the degradation caused by rain streaks in captured images, and existing imaging devices struggle to address this issue in real-time. While most efforts leverage deep networks for image deraining and have made progress, their large parameter sizes hinder deployment on resource-constrained devices. Additionally, these data-driven models often produce deterministic results, without considering their inherent epistemic uncertainty, which can lead to undesired reconstruction errors. Well-calibrated uncertainty can help alleviate prediction errors and assist measurement devices in mitigating risks and improving usability. Therefore, we propose an Uncertainty-Driven Multi-Scale Feature Fusion Network (UMFFNet) that learns the probability mapping distribution between paired images to estimate uncertainty. Specifically, we introduce an uncertainty feature fusion block (UFFB) that utilizes uncertainty information to dynamically enhance acquired features and focus on blurry regions obscured by rain streaks, reducing prediction errors. In addition, to further boost the performance of UMFFNet, we fused feature information from multiple scales to guide the network for efficient collaborative rain removal. Extensive experiments demonstrate that UMFFNet achieves significant performance improvements with few parameters, surpassing state-of-the-art image deraining methods.
Authors: Xiaohong Liu, Xiongkuo Min, Wei Sun, Yulun Zhang, Kai Zhang, Radu Timofte, Guangtao Zhai, Yixuan Gao, Yuqin Cao, Tengchuan Kou, Yunlong Dong, Ziheng Jia, Yilin Li, Wei Wu, Shuming Hu, Sibin Deng, Pengxiang Xiao, Ying Chen, Kai Li, Kai Zhao, Kun Yuan, Ming Sun, Heng Cong, Hao Wang, Lingzhi Fu, Yusheng Zhang, Rongyu Zhang, Hang Shi, Qihang Xu, Longan Xiao, Zhiliang Ma, Mirko Agarla, Luigi Celona, Claudio Rota, Raimondo Schettini, Zhiwei Huang, Yanan Li, Xiaotao Wang, Lei Lei, Hongye Liu, Wei Hong, Ironhead Chuang, Allen Lin, Drake Guan, Iris Chen, Kae Lou, Willy Huang, Yachun Tasi, Yvonne Kao, Haotian Fan, Fangyuan Kong, Shiqi Zhou, Hao Liu, Yu Lai, Shanshan Chen, Wenqi Wang, Haoning Wu, Chaofeng Chen, Chunzheng Zhu, Zekun Guo, Shiling Zhao, Haibing Yin, Hongkui Wang, Hanene Brachemi Meftah, et al. (8 additional authors not shown)
This paper reports on the NTIRE 2023 Quality Assessment of Video Enhancement Challenge, which will be held in conjunction with the New Trends in Image Restoration and Enhancement Workshop (NTIRE) at CVPR 2023. This challenge is to address a major challenge in the field of video processing, namely, video quality assessment (VQA) for enhanced videos. The challenge uses the VQA Dataset for Perceptual Video Enhancement (VDPVE), which has a total of 1211 enhanced videos, including 600 videos with color, brightness, and contrast enhancements, 310 videos with deblurring, and 301 deshaked videos. The challenge has a total of 167 registered participants. 61 participating teams submitted their prediction results during the development phase, with a total of 3168 submissions. A total of 176 submissions were submitted by 37 participating teams during the final testing phase. Finally, 19 participating teams submitted their models and fact sheets, and detailed the methods they used. Some methods have achieved better results than baseline methods, and the winning methods have demonstrated superior prediction performance.
Authors: Leyao Liu, Tao Kong, Minzhao Zhu, Jiashuo Fan, Lu Fang
3D instance segmentation methods often require fully-annotated dense labels for training, which are costly to obtain. In this paper, we present ClickSeg, a novel click-level weakly supervised 3D instance segmentation method that requires one point per instance annotation merely. Such a problem is very challenging due to the extremely limited labels, which has rarely been solved before. We first develop a baseline weakly-supervised training method, which generates pseudo labels for unlabeled data by the model itself. To utilize the property of click-level annotation setting, we further propose a new training framework. Instead of directly using the model inference way, i.e., mean-shift clustering, to generate the pseudo labels, we propose to use k-means with fixed initial seeds: the annotated points. New similarity metrics are further designed for clustering. Experiments on ScanNetV2 and S3DIS datasets show that the proposed ClickSeg surpasses the previous best weakly supervised instance segmentation result by a large margin (e.g., +9.4% mAP on ScanNetV2). Using 0.02% supervision signals merely, ClickSeg achieves $\sim$90% of the accuracy of the fully-supervised counterpart. Meanwhile, it also achieves state-of-the-art semantic segmentation results among weakly supervised methods that use the same annotation settings.
Authors: Ganlong Zhao, Guanbin Li, Yipeng Qin, Yizhou Yu
Dataset Condensation aims to condense a large dataset into a smaller one while maintaining its ability to train a well-performing model, thus reducing the storage cost and training effort in deep learning applications. However, conventional dataset condensation methods are optimization-oriented and condense the dataset by performing gradient or parameter matching during model optimization, which is computationally intensive even on small datasets and models. In this paper, we propose a novel dataset condensation method based on distribution matching, which is more efficient and promising. Specifically, we identify two important shortcomings of naive distribution matching (i.e., imbalanced feature numbers and unvalidated embeddings for distance computation) and address them with three novel techniques (i.e., partitioning and expansion augmentation, efficient and enriched model sampling, and class-aware distribution regularization). Our simple yet effective method outperforms most previous optimization-oriented methods with much fewer computational resources, thereby scaling data condensation to larger datasets and models. Extensive experiments demonstrate the effectiveness of our method. Codes are available at https://github.com/uitrbn/IDM
Authors: Feiran Hu, Peng Wang, Yangyang Li, Chenlong Duan, Zijian Zhu, Fei Wang, Faen Zhang, Yong Li, Xiu-Shen Wei
The SnakeCLEF2023 competition aims to the development of advanced algorithms for snake species identification through the analysis of images and accompanying metadata. This paper presents a method leveraging utilization of both images and metadata. Modern CNN models and strong data augmentation are utilized to learn better representation of images. To relieve the challenge of long-tailed distribution, seesaw loss is utilized in our method. We also design a light model to calculate prior probabilities using metadata features extracted from CLIP in post processing stage. Besides, we attach more importance to venomous species by assigning venomous species labels to some examples that model is uncertain about. Our method achieves 91.31% score of the final metric combined of F1 and other metrics on private leaderboard, which is the 1st place among the participators. The code is available at https://github.com/xiaoxsparraw/CLEF2023.
Authors: Hang Guo, Tao Dai, Guanghao Meng, Shu-Tao Xia
Scene text image super-resolution (STISR), aiming to improve image quality while boosting downstream scene text recognition accuracy, has recently achieved great success. However, most existing methods treat the foreground (character regions) and background (non-character regions) equally in the forward process, and neglect the disturbance from the complex background, thus limiting the performance. To address these issues, in this paper, we propose a novel method LEMMA that explicitly models character regions to produce high-level text-specific guidance for super-resolution. To model the location of characters effectively, we propose the location enhancement module to extract character region features based on the attention map sequence. Besides, we propose the multi-modal alignment module to perform bidirectional visual-semantic alignment to generate high-quality prior guidance, which is then incorporated into the super-resolution branch in an adaptive manner using the proposed adaptive fusion module. Experiments on TextZoom and four scene text recognition benchmarks demonstrate the superiority of our method over other state-of-the-art methods. Code is available at https://github.com/csguoh/LEMMA.
Authors: Changqi Wang, Haoyu Xie, Yuhui Yuan, Chong Fu, Xiangyu Yue
Semi-Supervised Semantic Segmentation (S4) aims to train a segmentation model with limited labeled images and a substantial volume of unlabeled images. To improve the robustness of representations, powerful methods introduce a pixel-wise contrastive learning approach in latent space (i.e., representation space) that aggregates the representations to their prototypes in a fully supervised manner. However, previous contrastive-based S4 methods merely rely on the supervision from the model's output (logits) in logit space during unlabeled training. In contrast, we utilize the outputs in both logit space and representation space to obtain supervision in a collaborative way. The supervision from two spaces plays two roles: 1) reduces the risk of over-fitting to incorrect semantic information in logits with the help of representations; 2) enhances the knowledge exchange between the two spaces. Furthermore, unlike previous approaches, we use the similarity between representations and prototypes as a new indicator to tilt training those under-performing representations and achieve a more efficient contrastive learning process. Results on two public benchmarks demonstrate the competitive performance of our method compared with state-of-the-art methods.
Authors: Yuzhong Zhao, Qixiang Ye, Weijia Wu, Chunhua Shen, Fang Wan
Weakly supervised object localization (WSOL) remains challenging when learning object localization models from image category labels. Conventional methods that discriminatively train activation models ignore representative yet less discriminative object parts. In this study, we propose a generative prompt model (GenPromp), defining the first generative pipeline to localize less discriminative object parts by formulating WSOL as a conditional image denoising procedure. During training, GenPromp converts image category labels to learnable prompt embeddings which are fed to a generative model to conditionally recover the input image with noise and learn representative embeddings. During inference, enPromp combines the representative embeddings with discriminative embeddings (queried from an off-the-shelf vision-language model) for both representative and discriminative capacity. The combined embeddings are finally used to generate multi-scale high-quality attention maps, which facilitate localizing full object extent. Experiments on CUB-200-2011 and ILSVRC show that GenPromp respectively outperforms the best discriminative models by 5.2% and 5.6% (Top-1 Loc), setting a solid baseline for WSOL with the generative model. Code is available at https://github.com/callsys/GenPromp.
Authors: Aaron Nicolson, Jason Dowling, Bevan Koopman
Chest X-Ray (CXR) report generation is a promising approach to improving the efficiency of CXR interpretation. However, a significant increase in diagnostic accuracy is required before that can be realised. Motivated by this, we propose a framework that is more inline with a radiologist's workflow by considering longitudinal data. Here, the decoder is additionally conditioned on the report from the subject's previous imaging study via a prompt. We also propose a new reward for reinforcement learning based on CXR-BERT, which computes the similarity between reports. We conduct experiments on the MIMIC-CXR dataset. The results indicate that longitudinal data improves CXR report generation. CXR-BERT is also shown to be a promising alternative to the current state-of-the-art reward based on RadGraph. This investigation indicates that longitudinal CXR report generation can offer a substantial increase in diagnostic accuracy. Our Hugging Face model is available at: https://huggingface.co/aehrc/cxrmate and code is available at: https://github.com/aehrc/cxrmate.
Authors: Qingwen Bu, Dong Huang, Heming Cui
The vulnerability of deep neural networks to adversarial samples has been a major impediment to their broad applications, despite their success in various fields. Recently, some works suggested that adversarially-trained models emphasize the importance of low-frequency information to achieve higher robustness. While several attempts have been made to leverage this frequency characteristic, they have all faced the issue that applying low-pass filters directly to input images leads to irreversible loss of discriminative information and poor generalizability to datasets with distinct frequency features. This paper presents a plug-and-play module called the Frequency Preference Control Module that adaptively reconfigures the low- and high-frequency components of intermediate feature representations, providing better utilization of frequency in robust learning. Empirical studies show that our proposed module can be easily incorporated into any adversarial training framework, further improving model robustness across different architectures and datasets. Additionally, experiments were conducted to examine how the frequency bias of robust models impacts the adversarial training process and its final robustness, revealing interesting insights.
Authors: Qinji Yu, Nan Xi, Junsong Yuan, Ziyu Zhou, Kang Dang, Xiaowei Ding
Unsupervised domain adaptation (UDA) has increasingly gained interests for its capacity to transfer the knowledge learned from a labeled source domain to an unlabeled target domain. However, typical UDA methods require concurrent access to both the source and target domain data, which largely limits its application in medical scenarios where source data is often unavailable due to privacy concern. To tackle the source data-absent problem, we present a novel two-stage source-free domain adaptation (SFDA) framework for medical image segmentation, where only a well-trained source segmentation model and unlabeled target data are available during domain adaptation. Specifically, in the prototype-anchored feature alignment stage, we first utilize the weights of the pre-trained pixel-wise classifier as source prototypes, which preserve the information of source features. Then, we introduce the bi-directional transport to align the target features with class prototypes by minimizing its expected cost. On top of that, a contrastive learning stage is further devised to utilize those pixels with unreliable predictions for a more compact target feature distribution. Extensive experiments on a cross-modality medical segmentation task demonstrate the superiority of our method in large domain discrepancy settings compared with the state-of-the-art SFDA approaches and even some UDA methods. Code is available at https://github.com/CSCYQJ/MICCAI23-ProtoContra-SFDA.
Authors: Xinyang Zhang, Wentian Zhao, Xin Lu, Jeff Chien
Layer compositing is one of the most popular image editing workflows among both amateurs and professionals. Motivated by the success of diffusion models, we explore layer compositing from a layered image generation perspective. Instead of generating an image, we propose to generate background, foreground, layer mask, and the composed image simultaneously. To achieve layered image generation, we train an autoencoder that is able to reconstruct layered images and train diffusion models on the latent representation. One benefit of the proposed problem is to enable better compositing workflows in addition to the high-quality image output. Another benefit is producing higher-quality layer masks compared to masks produced by a separate step of image segmentation. Experimental results show that the proposed method is able to generate high-quality layered images and initiates a benchmark for future work.
Authors: Along He, Kai Wang, Zhihong Wang, Tao Li, Huazhu Fu
Limited labeled data makes it hard to train models from scratch in medical domain, and an important paradigm is pre-training and then fine-tuning. Large pre-trained models contain rich representations, which can be adapted to downstream medical tasks. However, existing methods either tune all the parameters or the task-specific layers of the pre-trained models, ignoring the input variations of medical images, and thus they are not efficient or effective. In this work, we aim to study parameter-efficient fine-tuning (PEFT) for medical image analysis, and propose a dynamic visual prompt tuning method, named DVPT. It can extract knowledge beneficial to downstream tasks from large models with a few trainable parameters. Firstly, the frozen features are transformed by an lightweight bottleneck layer to learn the domain-specific distribution of downstream medical tasks, and then a few learnable visual prompts are used as dynamic queries and then conduct cross-attention with the transformed features, attempting to acquire sample-specific knowledge that are suitable for each sample. Finally, the features are projected to original feature dimension and aggregated with the frozen features. This DVPT module can be shared between different Transformer layers, further reducing the trainable parameters. To validate DVPT, we conduct extensive experiments with different pre-trained models on medical classification and segmentation tasks. We find such PEFT method can not only efficiently adapt the pre-trained models to the medical domain, but also brings data efficiency with partial labeled data. For example, with 0.5\% extra trainable parameters, our method not only outperforms state-of-the-art PEFT methods, even surpasses the full fine-tuning by more than 2.20\% Kappa score on medical classification task. It can saves up to 60\% labeled data and 99\% storage cost of ViT-B/16.
Authors: Quan Liu, Hongzi Zhu, Yunsong Zhou, Hongyang Li, Shan Chang, Minyi Guo
Registration of distant outdoor LiDAR point clouds is crucial to extending the 3D vision of collaborative autonomous vehicles, and yet is challenging due to small overlapping area and a huge disparity between observed point densities. In this paper, we propose Group-wise Contrastive Learning (GCL) scheme to extract density-invariant geometric features to register distant outdoor LiDAR point clouds. We mark through theoretical analysis and experiments that, contrastive positives should be independent and identically distributed (i.i.d.), in order to train densityinvariant feature extractors. We propose upon the conclusion a simple yet effective training scheme to force the feature of multiple point clouds in the same spatial location (referred to as positive groups) to be similar, which naturally avoids the sampling bias introduced by a pair of point clouds to conform with the i.i.d. principle. The resulting fully-convolutional feature extractor is more powerful and density-invariant than state-of-the-art methods, improving the registration recall of distant scenarios on KITTI and nuScenes benchmarks by 40.9% and 26.9%, respectively. The code will be open-sourced.
Authors: Zhenghao Feng, Lu Wen, Peng Wang, Binyu Yan, Xi Wu, Jiliu Zhou, Yan Wang
Currently, deep learning (DL) has achieved the automatic prediction of dose distribution in radiotherapy planning, enhancing its efficiency and quality. However, existing methods suffer from the over-smoothing problem for their commonly used L_1 or L_2 loss with posterior average calculations. To alleviate this limitation, we innovatively introduce a diffusion-based dose prediction (DiffDP) model for predicting the radiotherapy dose distribution of cancer patients. Specifically, the DiffDP model contains a forward process and a reverse process. In the forward process, DiffDP gradually transforms dose distribution maps into Gaussian noise by adding small noise and trains a noise predictor to predict the noise added in each timestep. In the reverse process, it removes the noise from the original Gaussian noise in multiple steps with the well-trained noise predictor and finally outputs the predicted dose distribution map. To ensure the accuracy of the prediction, we further design a structure encoder to extract anatomical information from patient anatomy images and enable the noise predictor to be aware of the dose constraints within several essential organs, i.e., the planning target volume and organs at risk. Extensive experiments on an in-house dataset with 130 rectum cancer patients demonstrate the s
Authors: Charilaos Papaioannou, Emmanouil Benetos, Alexandros Potamianos
Recent developments in MIR have led to several benchmark deep learning models whose embeddings can be used for a variety of downstream tasks. At the same time, the vast majority of these models have been trained on Western pop/rock music and related styles. This leads to research questions on whether these models can be used to learn representations for different music cultures and styles, or whether we can build similar music audio embedding models trained on data from different cultures or styles. To that end, we leverage transfer learning methods to derive insights about the similarities between the different music cultures to which the data belongs to. We use two Western music datasets, two traditional/folk datasets coming from eastern Mediterranean cultures, and two datasets belonging to Indian art music. Three deep audio embedding models are trained and transferred across domains, including two CNN-based and a Transformer-based architecture, to perform auto-tagging for each target domain dataset. Experimental results show that competitive performance is achieved in all domains via transfer learning, while the best source dataset varies for each music culture. The implementation and the trained models are both provided in a public repository.
Authors: Julia Grabinski, Janis Keuper, Margret Keuper
Convolutional neural networks encode images through a sequence of convolutions, normalizations and non-linearities as well as downsampling operations into potentially strong semantic embeddings. Yet, previous work showed that even slight mistakes during sampling, leading to aliasing, can be directly attributed to the networks' lack in robustness. To address such issues and facilitate simpler and faster adversarial training, [12] recently proposed FLC pooling, a method for provably alias-free downsampling - in theory. In this work, we conduct a further analysis through the lens of signal processing and find that such current pooling methods, which address aliasing in the frequency domain, are still prone to spectral leakage artifacts. Hence, we propose aliasing and spectral artifact-free pooling, short ASAP. While only introducing a few modifications to FLC pooling, networks using ASAP as downsampling method exhibit higher native robustness against common corruptions, a property that FLC pooling was missing. ASAP also increases native robustness against adversarial attacks on high and low resolution data while maintaining similar clean accuracy or even outperforming the baseline.
Authors: Xia Huang, Kai Fong Ernest Chong
Web image datasets curated online inherently contain ambiguous in-distribution (ID) instances and out-of-distribution (OOD) instances, which we collectively call non-conforming (NC) instances. In many recent approaches for mitigating the negative effects of NC instances, the core implicit assumption is that the NC instances can be found via entropy maximization. For "entropy" to be well-defined, we are interpreting the output prediction vector of an instance as the parameter vector of a multinomial random variable, with respect to some trained model with a softmax output layer. Hence, entropy maximization is based on the idealized assumption that NC instances have predictions that are "almost" uniformly distributed. However, in real-world web image datasets, there are numerous NC instances whose predictions are far from being uniformly distributed. To tackle the limitation of entropy maximization, we propose $(\alpha, \beta)$-generalized KL divergence, $\mathcal{D}_{\text{KL}}^{\alpha, \beta}(p\|q)$, which can be used to identify significantly more NC instances. Theoretical properties of $\mathcal{D}_{\text{KL}}^{\alpha, \beta}(p\|q)$ are proven, and we also show empirically that a simple use of $\mathcal{D}_{\text{KL}}^{\alpha, \beta}(p\|q)$ outperforms all baselines on the NC instance identification task. Building upon $(\alpha,\beta)$-generalized KL divergence, we also introduce a new iterative training framework, GenKL, that identifies and relabels NC instances. When evaluated on three web image datasets, Clothing1M, Food101/Food101N, and mini WebVision 1.0, we achieved new state-of-the-art classification accuracies: $81.34\%$, $85.73\%$ and $78.99\%$/$92.54\%$ (top-1/top-5), respectively.
Authors: Hao Yang, Liyuan Pan, Yan Yang, Miaomiao Liu,
Recovering sharp images from dual-pixel (DP) pairs with disparity-dependent blur is a challenging task.~Existing blur map-based deblurring methods have demonstrated promising results. In this paper, we propose, to the best of our knowledge, the first framework to introduce the contrastive language-image pre-training framework (CLIP) to achieve accurate blur map estimation from DP pairs unsupervisedly. To this end, we first carefully design text prompts to enable CLIP to understand blur-related geometric prior knowledge from the DP pair. Then, we propose a format to input stereo DP pair to the CLIP without any fine-tuning, where the CLIP is pre-trained on monocular images. Given the estimated blur map, we introduce a blur-prior attention block, a blur-weighting loss and a blur-aware loss to recover the all-in-focus image. Our method achieves state-of-the-art performance in extensive experiments.
Authors: Yinghao Zhang, Xiaodi Li, Weihang Li, Yue Hu
Deep unrolling networks that utilize sparsity priors have achieved great success in dynamic magnetic resonance (MR) imaging. The convolutional neural network (CNN) is usually utilized to extract the transformed domain, and then the soft thresholding (ST) operator is applied to the CNN-transformed data to enforce the sparsity priors. However, the ST operator is usually constrained to be the same across all channels of the CNN-transformed data. In this paper, we propose a novel operator, called soft thresholding with channel attention (AST), that learns the threshold for each channel. In particular, we put forward a novel deep unrolling shrinkage network (DUS-Net) by unrolling the alternating direction method of multipliers (ADMM) for optimizing the transformed $l_1$ norm dynamic MR reconstruction model. Experimental results on an open-access dynamic cine MR dataset demonstrate that the proposed DUS-Net outperforms the state-of-the-art methods. The source code is available at \url{https://github.com/yhao-z/DUS-Net}.
Authors: Zhigang Chang, Weitai Hu, Qing Yang, Shibao Zheng
In dyadic speaker-listener interactions, the listener's head reactions along with the speaker's head movements, constitute an important non-verbal semantic expression together. The listener Head generation task aims to synthesize responsive listener's head videos based on audios of the speaker and reference images of the listener. Compared to the Talking-head generation, it is more challenging to capture the correlation clues from the speaker's audio and visual information. Following the ViCo baseline scheme, we propose a high-performance solution by enhancing the hierarchical semantic extraction capability of the audio encoder module and improving the decoder part, renderer and post-processing modules. Our solution gets the first place on the official leaderboard for the track of listening head generation. This paper is a technical report of ViCo@2023 Conversational Head Generation Challenge in ACM Multimedia 2023 conference.
Authors: Lydia Abady, Jun Wang, Benedetta Tondi, Mauro Barni
Despite the wide variety of methods developed for synthetic image attribution, most of them can only attribute images generated by models or architectures included in the training set and do not work with unknown architectures, hindering their applicability in real-world scenarios. In this paper, we propose a verification framework that relies on a Siamese Network to address the problem of open-set attribution of synthetic images to the architecture that generated them. We consider two different settings. In the first setting, the system determines whether two images have been produced by the same generative architecture or not. In the second setting, the system verifies a claim about the architecture used to generate a synthetic image, utilizing one or multiple reference images generated by the claimed architecture. The main strength of the proposed system is its ability to operate in both closed and open-set scenarios so that the input images, either the query and reference images, can belong to the architectures considered during training or not. Experimental evaluations encompassing various generative architectures such as GANs, diffusion models, and transformers, focusing on synthetic face image generation, confirm the excellent performance of our method in both closed and open-set settings, as well as its strong generalization capabilities.
Authors: Yaran Chen, Xueyu Chen, Yu Han, Haoran Li, Dongbin Zhao, Jingzhong Li, Xu Wang
Non alcoholic fatty liver disease (NAFLD) is the most common cause of chronic liver disease, which can be predicted accurately to prevent advanced fibrosis and cirrhosis. While, a liver biopsy, the gold standard for NAFLD diagnosis, is invasive, expensive, and prone to sampling errors. Therefore, non-invasive studies are extremely promising, yet they are still in their infancy due to the lack of comprehensive research data and intelligent methods for multi-modal data. This paper proposes a NAFLD diagnosis system (DeepFLDDiag) combining a comprehensive clinical dataset (FLDData) and a multi-modal learning based NAFLD prediction method (DeepFLD). The dataset includes over 6000 participants physical examinations, laboratory and imaging studies, extensive questionnaires, and facial images of partial participants, which is comprehensive and valuable for clinical studies. From the dataset, we quantitatively analyze and select clinical metadata that most contribute to NAFLD prediction. Furthermore, the proposed DeepFLD, a deep neural network model designed to predict NAFLD using multi-modal input, including metadata and facial images, outperforms the approach that only uses metadata. Satisfactory performance is also verified on other unseen datasets. Inspiringly, DeepFLD can achieve competitive results using only facial images as input rather than metadata, paving the way for a more robust and simpler non-invasive NAFLD diagnosis.
Authors: Umberto Michieli, Mete Ozay
Vision systems mounted on home robots need to interact with unseen classes in changing environments. Robots have limited computational resources, labelled data and storage capability. These requirements pose some unique challenges: models should adapt without forgetting past knowledge in a data- and parameter-efficient way. We characterize the problem as few-shot (FS) online continual learning (OCL), where robotic agents learn from a non-repeated stream of few-shot data updating only a few model parameters. Additionally, such models experience variable conditions at test time, where objects may appear in different poses (e.g., horizontal or vertical) and environments (e.g., day or night). To improve robustness of CL agents, we propose RobOCLe, which; 1) constructs an enriched feature space computing high order statistical moments from the embedded features of samples; and 2) computes similarity between high order statistics of the samples on the enriched feature space, and predicts their class labels. We evaluate robustness of CL models to train/test augmentations in various cases. We show that different moments allow RobOCLe to capture different properties of deformations, providing higher robustness with no decrease of inference speed.
Authors: Shunxin Wang, Raymond Veldhuis, Christoph Brune, Nicola Strisciuglio
Frequency analysis is useful for understanding the mechanisms of representation learning in neural networks (NNs). Most research in this area focuses on the learning dynamics of NNs for regression tasks, while little for classification. This study empirically investigates the latter and expands the understanding of frequency shortcuts. First, we perform experiments on synthetic datasets, designed to have a bias in different frequency bands. Our results demonstrate that NNs tend to find simple solutions for classification, and what they learn first during training depends on the most distinctive frequency characteristics, which can be either low- or high-frequencies. Second, we confirm this phenomenon on natural images. We propose a metric to measure class-wise frequency characteristics and a method to identify frequency shortcuts. The results show that frequency shortcuts can be texture-based or shape-based, depending on what best simplifies the objective. Third, we validate the transferability of frequency shortcuts on out-of-distribution (OOD) test sets. Our results suggest that frequency shortcuts can be transferred across datasets and cannot be fully avoided by larger model capacity and data augmentation. We recommend that future research should focus on effective training schemes mitigating frequency shortcut learning.
Authors: Ajay Gunalan, Marco Castello, Simonluca Piazza, Shunlei Li, Alberto Diaspro, Leonardo S. Mattos, Paolo Bianchini
We present a novel approach to implement compressive sensing in laser scanning microscopes (LSM), specifically in image scanning microscopy (ISM), using a single-photon avalanche diode (SPAD) array detector. Our method addresses two significant limitations in applying compressive sensing to LSM: the time to compute the sampling matrix and the quality of reconstructed images. We employ a fixed sampling strategy, skipping alternate rows and columns during data acquisition, which reduces the number of points scanned by a factor of four and eliminates the need to compute different sampling matrices. By exploiting the parallel images generated by the SPAD array, we improve the quality of the reconstructed compressive-ISM images compared to standard compressive confocal LSM images. Our results demonstrate the effectiveness of our approach in producing higher-quality images with reduced data acquisition time and potential benefits in reducing photobleaching.
Authors: Szu-Chi Chung
In single-particle cryo-electron microscopy (cryo-EM), the efficient determination of orientation parameters for 2D projection images poses a significant challenge yet is crucial for reconstructing 3D structures. This task is complicated by the high noise levels present in the cryo-EM datasets, which often include outliers, necessitating several time-consuming 2D clean-up processes. Recently, solutions based on deep learning have emerged, offering a more streamlined approach to the traditionally laborious task of orientation estimation. These solutions often employ amortized inference, eliminating the need to estimate parameters individually for each image. However, these methods frequently overlook the presence of outliers and may not adequately concentrate on the components used within the network. This paper introduces a novel approach that uses a 10-dimensional feature vector to represent the orientation and applies a Quadratically-Constrained Quadratic Program to derive the predicted orientation as a unit quaternion, supplemented by an uncertainty metric. Furthermore, we propose a unique loss function that considers the pairwise distances between orientations, thereby enhancing the accuracy of our method. Finally, we also comprehensively evaluate the design choices involved in constructing the encoder network, a topic that has not received sufficient attention in the literature. Our numerical analysis demonstrates that our methodology effectively recovers orientations from 2D cryo-EM images in an end-to-end manner. Importantly, the inclusion of uncertainty quantification allows for direct clean-up of the dataset at the 3D level. Lastly, we package our proposed methods into a user-friendly software suite named cryo-forum, designed for easy accessibility by the developers.
Authors: Lei Wang, Bo Liu, Fangfang Liang, Bincheng Wang
Gait recognition is a biometric technique that identifies individuals by their unique walking styles, which is suitable for unconstrained environments and has a wide range of applications. While current methods focus on exploiting body part-based representations, they often neglect the hierarchical dependencies between local motion patterns. In this paper, we propose a hierarchical spatio-temporal representation learning (HSTL) framework for extracting gait features from coarse to fine. Our framework starts with a hierarchical clustering analysis to recover multi-level body structures from the whole body to local details. Next, an adaptive region-based motion extractor (ARME) is designed to learn region-independent motion features. The proposed HSTL then stacks multiple ARMEs in a top-down manner, with each ARME corresponding to a specific partition level of the hierarchy. An adaptive spatio-temporal pooling (ASTP) module is used to capture gait features at different levels of detail to perform hierarchical feature mapping. Finally, a frame-level temporal aggregation (FTA) module is employed to reduce redundant information in gait sequences through multi-scale temporal downsampling. Extensive experiments on CASIA-B, OUMVLP, GREW, and Gait3D datasets demonstrate that our method outperforms the state-of-the-art while maintaining a reasonable balance between model accuracy and complexity.
Authors: Hassan Khalid, Nisar Ahmed
BIQA (Blind Image Quality Assessment) is an important field of study that evaluates images automatically. Although significant progress has been made, blind image quality assessment remains a difficult task since images vary in content and distortions. Most algorithms generate quality without emphasizing the important region of interest. In order to solve this, a multi-stream spatial and channel attention-based algorithm is being proposed. This algorithm generates more accurate predictions with a high correlation to human perceptual assessment by combining hybrid features from two different backbones, followed by spatial and channel attention to provide high weights to the region of interest. Four legacy image quality assessment datasets are used to validate the effectiveness of our proposed approach. Authentic and synthetic distortion image databases are used to demonstrate the effectiveness of the proposed method, and we show that it has excellent generalization properties with a particular focus on the perceptual foreground information.
Authors: Jitao Ma, Weiying Xie, Yunsong Li, Leyuan Fang
Hyperspectral anomaly detection (HAD) is widely used in Earth observation and deep space exploration. A major challenge for HAD is the complex background of the input hyperspectral images (HSIs), resulting in anomalies confused in the background. On the other hand, the lack of labeled samples for HSIs leads to poor generalization of existing HAD methods. This paper starts the first attempt to study a new and generalizable background learning problem without labeled samples. We present a novel solution BSDM (background suppression diffusion model) for HAD, which can simultaneously learn latent background distributions and generalize to different datasets for suppressing complex background. It is featured in three aspects: (1) For the complex background of HSIs, we design pseudo background noise and learn the potential background distribution in it with a diffusion model (DM). (2) For the generalizability problem, we apply a statistical offset module so that the BSDM adapts to datasets of different domains without labeling samples. (3) For achieving background suppression, we innovatively improve the inference process of DM by feeding the original HSIs into the denoising network, which removes the background as noise. Our work paves a new background suppression way for HAD that can improve HAD performance without the prerequisite of manually labeled data. Assessments and generalization experiments of four HAD methods on several real HSI datasets demonstrate the above three unique properties of the proposed method. The code is available at https://github.com/majitao-xd/BSDM-HAD.
Authors: Liekang Zeng, Haowei Chen, Daipeng Feng, Xiaoxi Zhang, Xu Chen
Accurate navigation is of paramount importance to ensure flight safety and efficiency for autonomous drones. Recent research starts to use Deep Neural Networks to enhance drone navigation given their remarkable predictive capability for visual perception. However, existing solutions either run DNN inference tasks on drones in situ, impeded by the limited onboard resource, or offload the computation to external servers which may incur large network latency. Few works consider jointly optimizing the offloading decisions along with image transmission configurations and adapting them on the fly. In this paper, we propose A3D, an edge server assisted drone navigation framework that can dynamically adjust task execution location, input resolution, and image compression ratio in order to achieve low inference latency, high prediction accuracy, and long flight distances. Specifically, we first augment state-of-the-art convolutional neural networks for drone navigation and define a novel metric called Quality of Navigation as our optimization objective which can effectively capture the above goals. We then design a deep reinforcement learning based neural scheduler at the drone side for which an information encoder is devised to reshape the state features and thus improve its learning ability. To further support simultaneous multi-drone serving, we extend the edge server design by developing a network-aware resource allocation algorithm, which allows provisioning containerized resources aligned with drones' demand. We finally implement a proof-of-concept prototype with realistic devices and validate its performance in a real-world campus scene, as well as a simulation environment for thorough evaluation upon AirSim. Extensive experimental results show that A3D can reduce end-to-end latency by 28.06% and extend the flight distance by up to 27.28% compared with non-adaptive solutions.
Authors: Tatiana Fountoukidou, Raphael Sznitman
Recent advances in machine learning models have greatly increased the performance of automated methods in medical image analysis. However, the internal functioning of such models is largely hidden, which hinders their integration in clinical practice. Explainability and trust are viewed as important aspects of modern methods, for the latter's widespread use in clinical communities. As such, validation of machine learning models represents an important aspect and yet, most methods are only validated in a limited way. In this work, we focus on providing a richer and more appropriate validation approach for highly powerful Visual Question Answering (VQA) algorithms. To better understand the performance of these methods, which answer arbitrary questions related to images, this work focuses on an automatic visual Turing test (VTT). That is, we propose an automatic adaptive questioning method, that aims to expose the reasoning behavior of a VQA algorithm. Specifically, we introduce a reinforcement learning (RL) agent that observes the history of previously asked questions, and uses it to select the next question to pose. We demonstrate our approach in the context of evaluating algorithms that automatically answer questions related to diabetic macular edema (DME) grading. The experiments show that such an agent has similar behavior to a clinician, whereby asking questions that are relevant to key clinical concepts.
Authors: Hao Su, Xuefeng Liu, Jianwei Niu, Ji Wan, Xinghao Wu
We propose 3Deformer, a general-purpose framework for interactive 3D shape editing. Given a source 3D mesh with semantic materials, and a user-specified semantic image, 3Deformer can accurately edit the source mesh following the shape guidance of the semantic image, while preserving the source topology as rigid as possible. Recent studies of 3D shape editing mostly focus on learning neural networks to predict 3D shapes, which requires high-cost 3D training datasets and is limited to handling objects involved in the datasets. Unlike these studies, our 3Deformer is a non-training and common framework, which only requires supervision of readily-available semantic images, and is compatible with editing various objects unlimited by datasets. In 3Deformer, the source mesh is deformed utilizing the differentiable renderer technique, according to the correspondences between semantic images and mesh materials. However, guiding complex 3D shapes with a simple 2D image incurs extra challenges, that is, the deform accuracy, surface smoothness, geometric rigidity, and global synchronization of the edited mesh should be guaranteed. To address these challenges, we propose a hierarchical optimization architecture to balance the global and local shape features, and propose further various strategies and losses to improve properties of accuracy, smoothness, rigidity, and so on. Extensive experiments show that our 3Deformer is able to produce impressive results and reaches the state-of-the-art level.
Authors: Saad Manzur, Wayne Hayes
Existing 2D-to-3D pose lifting networks suffer from poor performance in cross-dataset benchmarks. Although the use of 2D keypoints joined by "stick-figure" limbs has shown promise as an intermediate step, stick-figures do not account for occlusion information that is often inherent in an image. In this paper, we propose a novel representation using opaque 3D limbs that preserves occlusion information while implicitly encoding joint locations. Crucially, when training on data with accurate three-dimensional keypoints and without part-maps, this representation allows training on abstract synthetic images, with occlusion, from as many synthetic viewpoints as desired. The result is a pose defined by limb angles rather than joint positions $\unicode{x2013}$ because poses are, in the real world, independent of cameras $\unicode{x2013}$ allowing us to predict poses that are completely independent of camera viewpoint. The result provides not only an improvement in same-dataset benchmarks, but a "quantum leap" in cross-dataset benchmarks.
Authors: Fa-Ting Hong, Dan Xu
Talking head video generation aims to animate a human face in a still image with dynamic poses and expressions using motion information derived from a target-driving video, while maintaining the person's identity in the source image. However, dramatic and complex motions in the driving video cause ambiguous generation, because the still source image cannot provide sufficient appearance information for occluded regions or delicate expression variations, which produces severe artifacts and significantly degrades the generation quality. To tackle this problem, we propose to learn a global facial representation space, and design a novel implicit identity representation conditioned memory compensation network, coined as MCNet, for high-fidelity talking head generation.~Specifically, we devise a network module to learn a unified spatial facial meta-memory bank from all training samples, which can provide rich facial structure and appearance priors to compensate warped source facial features for the generation. Furthermore, we propose an effective query mechanism based on implicit identity representations learned from the discrete keypoints of the source image. It can greatly facilitate the retrieval of more correlated information from the memory bank for the compensation. Extensive experiments demonstrate that MCNet can learn representative and complementary facial memory, and can clearly outperform previous state-of-the-art talking head generation methods on VoxCeleb1 and CelebV datasets. Please check our \href{https://github.com/harlanhong/ICCV2023-MCNET}{Project}.
Authors: Zijie Song, Zhenzhen Hu, Richang Hong
Cross-lingual image captioning is confronted with both cross-lingual and cross-modal challenges for multimedia analysis. The crucial issue in this task is to model the global and local matching between the image and different languages. Existing cross-modal embedding methods based on Transformer architecture oversight the local matching between the image region and monolingual words, not to mention in the face of a variety of differentiated languages. Due to the heterogeneous property of the cross-modal and cross-lingual task, we utilize the heterogeneous network to establish cross-domain relationships and the local correspondences between the image and different languages. In this paper, we propose an Embedded Heterogeneous Attention Transformer (EHAT) to build reasoning paths bridging cross-domain for cross-lingual image captioning and integrate into transformer. The proposed EHAT consists of a Masked Heterogeneous Cross-attention (MHCA), Heterogeneous Attention Reasoning Network (HARN) and Heterogeneous Co-attention (HCA). HARN as the core network, models and infers cross-domain relationship anchored by vision bounding box representation features to connect two languages word features and learn the heterogeneous maps. MHCA and HCA implement cross-domain integration in the encoder through the special heterogeneous attention and enable single model to generate two language captioning. We test on MSCOCO dataset to generate English and Chinese, which are most widely used and have obvious difference between their language families. Our experiments show that our method even achieve better than advanced monolingual methods.
Authors: Mochu Xiang, Jing Zhang, Nick Barnes, Yuchao Dai
Effectively measuring and modeling the reliability of a trained model is essential to the real-world deployment of monocular depth estimation (MDE) models. However, the intrinsic ill-posedness and ordinal-sensitive nature of MDE pose major challenges to the estimation of uncertainty degree of the trained models. On the one hand, utilizing current uncertainty modeling methods may increase memory consumption and are usually time-consuming. On the other hand, measuring the uncertainty based on model accuracy can also be problematic, where uncertainty reliability and prediction accuracy are not well decoupled. In this paper, we propose to model the uncertainty of MDE models from the perspective of the inherent probability distributions originating from the depth probability volume and its extensions, and to assess it more fairly with more comprehensive metrics. By simply introducing additional training regularization terms, our model, with surprisingly simple formations and without requiring extra modules or multiple inferences, can provide uncertainty estimations with state-of-the-art reliability, and can be further improved when combined with ensemble or sampling methods. A series of experiments demonstrate the effectiveness of our methods.
Authors: Matteo Ronchetti, Wolfgang Wein, Nassir Navab, Oliver Zettinig, Raphael Prevost
Multimodal image registration is a challenging but essential step for numerous image-guided procedures. Most registration algorithms rely on the computation of complex, frequently non-differentiable similarity metrics to deal with the appearance discrepancy of anatomical structures between imaging modalities. Recent Machine Learning based approaches are limited to specific anatomy-modality combinations and do not generalize to new settings. We propose a generic framework for creating expressive cross-modal descriptors that enable fast deformable global registration. We achieve this by approximating existing metrics with a dot-product in the feature space of a small convolutional neural network (CNN) which is inherently differentiable can be trained without registered data. Our method is several orders of magnitude faster than local patch-based metrics and can be directly applied in clinical settings by replacing the similarity measure with the proposed one. Experiments on three different datasets demonstrate that our approach generalizes well beyond the training data, yielding a broad capture range even on unseen anatomies and modality pairs, without the need for specialized retraining. We make our training code and data publicly available.
Authors: Cian Eastwood, Shashank Singh, Andrei Liviu Nicolicioiu, Marin Vlastelica, Julius von Kügelgen, Bernhard Schölkopf
To avoid failures on out-of-distribution data, recent works have sought to extract features that have a stable or invariant relationship with the label across domains, discarding the "spurious" or unstable features whose relationship with the label changes across domains. However, unstable features often carry complementary information about the label that could boost performance if used correctly in the test domain. Our main contribution is to show that it is possible to learn how to use these unstable features in the test domain without labels. In particular, we prove that pseudo-labels based on stable features provide sufficient guidance for doing so, provided that stable and unstable features are conditionally independent given the label. Based on this theoretical insight, we propose Stable Feature Boosting (SFB), an algorithm for: (i) learning a predictor that separates stable and conditionally-independent unstable features; and (ii) using the stable-feature predictions to adapt the unstable-feature predictions in the test domain. Theoretically, we prove that SFB can learn an asymptotically-optimal predictor without test-domain labels. Empirically, we demonstrate the effectiveness of SFB on real and synthetic data.
Authors: Pedro Gomes, Silvia Rossi, Laura Toni
This paper focuses on motion prediction for point cloud sequences in the challenging case of deformable 3D objects, such as human body motion. First, we investigate the challenges caused by deformable shapes and complex motions present in this type of representation, with the ultimate goal of understanding the technical limitations of state-of-the-art models. From this understanding, we propose an improved architecture for point cloud prediction of deformable 3D objects. Specifically, to handle deformable shapes, we propose a graph-based approach that learns and exploits the spatial structure of point clouds to extract more representative features. Then we propose a module able to combine the learned features in an adaptative manner according to the point cloud movements. The proposed adaptative module controls the composition of local and global motions for each point, enabling the network to model complex motions in deformable 3D objects more effectively. We tested the proposed method on the following datasets: MNIST moving digits, the Mixamo human bodies motions, JPEG and CWIPC-SXR real-world dynamic bodies. Simulation results demonstrate that our method outperforms the current baseline methods given its improved ability to model complex movements as well as preserve point cloud shape. Furthermore, we demonstrate the generalizability of the proposed framework for dynamic feature learning, by testing the framework for action recognition on the MSRAction3D dataset and achieving results on-par with state-of-the-art methods
Authors: Miles Everett, Mingjun Zhong, Georgios Leontidis
Capsule Networks have emerged as a powerful class of deep learning architectures, known for robust performance with relatively few parameters compared to Convolutional Neural Networks (CNNs). However, their inherent efficiency is often overshadowed by their slow, iterative routing mechanisms which establish connections between Capsule layers, posing computational challenges resulting in an inability to scale. In this paper, we introduce a novel, non-iterative routing mechanism, inspired by trainable prototype clustering. This innovative approach aims to mitigate computational complexity, while retaining, if not enhancing, performance efficacy. Furthermore, we harness a shared Capsule subspace, negating the need to project each lower-level Capsule to each higher-level Capsule, thereby significantly reducing memory requisites during training. Our approach demonstrates superior results compared to the current best non-iterative Capsule Network and tests on the Imagewoof dataset, which is too computationally demanding to handle efficiently by iterative approaches. Our findings underscore the potential of our proposed methodology in enhancing the operational efficiency and performance of Capsule Networks, paving the way for their application in increasingly complex computational scenarios.
Authors: Steven Landgraf, Markus Hillemann, Kira Wursthorn, Markus Ulrich
Deep neural networks have shown exceptional performance in various tasks, but their lack of robustness, reliability, and tendency to be overconfident pose challenges for their deployment in safety-critical applications like autonomous driving. In this regard, quantifying the uncertainty inherent to a model's prediction is a promising endeavour to address these shortcomings. In this work, we present a novel Uncertainty-aware Cross-Entropy loss (U-CE) that incorporates dynamic predictive uncertainties into the training process by pixel-wise weighting of the well-known cross-entropy loss (CE). Through extensive experimentation, we demonstrate the superiority of U-CE over regular CE training on two benchmark datasets, Cityscapes and ACDC, using two common backbone architectures, ResNet-18 and ResNet-101. With U-CE, we manage to train models that not only improve their segmentation performance but also provide meaningful uncertainties after training. Consequently, we contribute to the development of more robust and reliable segmentation models, ultimately advancing the state-of-the-art in safety-critical applications and beyond.
Authors: Zuozhuo Dai, Fangtao Shao, Qingkun Su, Zilong Dong, Siyu Zhu
State-of-the-art text-video retrieval (TVR) methods typically utilize CLIP and cosine similarity for efficient retrieval. Meanwhile, cross attention methods, which employ a transformer decoder to compute attention between each text query and all frames in a video, offer a more comprehensive interaction between text and videos. However, these methods lack important fine-grained spatial information as they directly compute attention between text and video-level tokens. To address this issue, we propose CrossTVR, a two-stage text-video retrieval architecture. In the first stage, we leverage existing TVR methods with cosine similarity network for efficient text/video candidate selection. In the second stage, we propose a novel decoupled video text cross attention module to capture fine-grained multimodal information in spatial and temporal dimensions. Additionally, we employ the frozen CLIP model strategy in fine-grained retrieval, enabling scalability to larger pre-trained vision models like ViT-G, resulting in improved retrieval performance. Experiments on text video retrieval datasets demonstrate the effectiveness and scalability of our proposed CrossTVR compared to state-of-the-art approaches.
Authors: Longxiang Tang, Kai Li, Chunming He, Yulun Zhang, Xiu Li
This paper studies source-free domain adaptive fundus image segmentation which aims to adapt a pretrained fundus segmentation model to a target domain using unlabeled images. This is a challenging task because it is highly risky to adapt a model only using unlabeled data. Most existing methods tackle this task mainly by designing techniques to carefully generate pseudo labels from the model's predictions and use the pseudo labels to train the model. While often obtaining positive adaption effects, these methods suffer from two major issues. First, they tend to be fairly unstable - incorrect pseudo labels abruptly emerged may cause a catastrophic impact on the model. Second, they fail to consider the severe class imbalance of fundus images where the foreground (e.g., cup) region is usually very small. This paper aims to address these two issues by proposing the Class-Balanced Mean Teacher (CBMT) model. CBMT addresses the unstable issue by proposing a weak-strong augmented mean teacher learning scheme where only the teacher model generates pseudo labels from weakly augmented images to train a student model that takes strongly augmented images as input. The teacher is updated as the moving average of the instantly trained student, which could be noisy. This prevents the teacher model from being abruptly impacted by incorrect pseudo-labels. For the class imbalance issue, CBMT proposes a novel loss calibration approach to highlight foreground classes according to global statistics. Experiments show that CBMT well addresses these two issues and outperforms existing methods on multiple benchmarks.
Authors: Siyan Dong, Shaohui Liu, Hengkai Guo, Baoquan Chen, Marc Pollefeys
Visual (re)localization is critical for various applications in computer vision and robotics. Its goal is to estimate the 6 degrees of freedom (DoF) camera pose for each query image, based on a set of posed database images. Currently, all leading solutions are structure-based that either explicitly construct 3D metric maps from the database with structure-from-motion, or implicitly encode the 3D information with scene coordinate regression models. On the contrary, visual localization without reconstructing the scene in 3D offers clear benefits. It makes deployment more convenient by reducing database pre-processing time, releasing storage requirements, and remaining unaffected by imperfect reconstruction, etc. In this technical report, we demonstrate that it is possible to achieve high localization accuracy without reconstructing the scene from the database. The key to achieving this owes to a tailored motion averaging over database-query pairs. Experiments show that our visual localization proposal, LazyLoc, achieves comparable performance against state-of-the-art structure-based methods. Furthermore, we showcase the versatility of LazyLoc, which can be easily extended to handle complex configurations such as multi-query co-localization and camera rigs.
Authors: Young D. Kwon, Rui Li, Stylianos I. Venieris, Jagmohan Chauhan, Nicholas D. Lane, Cecilia Mascolo
On-device training is essential for user personalisation and privacy. With the pervasiveness of IoT devices and microcontroller units (MCU), this task becomes more challenging due to the constrained memory and compute resources, and the limited availability of labelled user data. Nonetheless, prior works neglect the data scarcity issue, require excessively long training time (e.g. a few hours), or induce substantial accuracy loss ($\geq$10\%). We propose TinyTrain, an on-device training approach that drastically reduces training time by selectively updating parts of the model and explicitly coping with data scarcity. TinyTrain introduces a task-adaptive sparse-update method that dynamically selects the layer/channel based on a multi-objective criterion that jointly captures user data, the memory, and the compute capabilities of the target device, leading to high accuracy on unseen tasks with reduced computation and memory footprint. TinyTrain outperforms vanilla fine-tuning of the entire network by 3.6-5.0\% in accuracy, while reducing the backward-pass memory and computation cost by up to 2,286$\times$ and 7.68$\times$, respectively. Targeting broadly used real-world edge devices, TinyTrain achieves 9.5$\times$ faster and 3.5$\times$ more energy-efficient training over status-quo approaches, and 2.8$\times$ smaller memory footprint than SOTA approaches, while remaining within the 1 MB memory envelope of MCU-grade platforms.
Authors: Carl Shneider, Peyman Rostami, Anis Kacem, Nilotpal Sinha, Abd El Rahman Shabayek, Djamila Aouada
Deploying deep learning neural networks on edge devices, to accomplish task specific objectives in the real-world, requires a reduction in their memory footprint, power consumption, and latency. This can be realized via efficient model compression. Disentangled latent representations produced by variational autoencoder (VAE) networks are a promising approach for achieving model compression because they mainly retain task-specific information, discarding useless information for the task at hand. We make use of the Beta-VAE framework combined with a standard criterion for pruning to investigate the impact of forcing the network to learn disentangled representations on the pruning process for the task of classification. In particular, we perform experiments on MNIST and CIFAR10 datasets, examine disentanglement challenges, and propose a path forward for future works.
Authors: Isabel Funke, Dominik Rivoir, Stefanie Krell, Stefanie Speidel
To enable context-aware computer assistance in the operating room of the future, cognitive systems need to understand automatically which surgical phase is being performed by the medical team. The primary source of information for surgical phase recognition is typically video, which presents two challenges: extracting meaningful features from the video stream and effectively modeling temporal information in the sequence of visual features. For temporal modeling, attention mechanisms have gained popularity due to their ability to capture long-range dependencies. In this paper, we explore design choices for attention in existing temporal models for surgical phase recognition and propose a novel approach that does not resort to local attention or regularization of attention weights: TUNeS is an efficient and simple temporal model that incorporates self-attention at the coarsest stage of a U-Net-like structure. In addition, we propose to train the feature extractor, a standard CNN, together with an LSTM on preferably long video segments, i.e., with long temporal context. In our experiments, all temporal models performed better on top of feature extractors that were trained with longer temporal context. On top of these contextualized features, TUNeS achieves state-of-the-art results on Cholec80.
Authors: Julia Grabinski, Janis Keuper, Margret Keuper
Motivated by the recent trend towards the usage of larger receptive fields for more context-aware neural networks in vision applications, we aim to investigate how large these receptive fields really need to be. To facilitate such study, several challenges need to be addressed, most importantly: (i) We need to provide an effective way for models to learn large filters (potentially as large as the input data) without increasing their memory consumption during training or inference, (ii) the study of filter sizes has to be decoupled from other effects such as the network width or number of learnable parameters, and (iii) the employed convolution operation should be a plug-and-play module that can replace any conventional convolution in a Convolutional Neural Network (CNN) and allow for an efficient implementation in current frameworks. To facilitate such models, we propose to learn not spatial but frequency representations of filter weights as neural implicit functions, such that even infinitely large filters can be parameterized by only a few learnable weights. The resulting neural implicit frequency CNNs are the first models to achieve results on par with the state-of-the-art on large image classification benchmarks while executing convolutions solely in the frequency domain and can be employed within any CNN architecture. They allow us to provide an extensive analysis of the learned receptive fields. Interestingly, our analysis shows that, although the proposed networks could learn very large convolution kernels, the learned filters practically translate into well-localized and relatively small convolution kernels in the spatial domain.
Authors: Amirhossein Aminimehr, Pouya Khani, Amirali Molaei, Amirmohammad Kazemeini, Erik Cambria
The field of Explainable Artificial Intelligence (XAI) aims to improve the interpretability of black-box machine learning models. Building a heatmap based on the importance value of input features is a popular method for explaining the underlying functions of such models in producing their predictions. Heatmaps are almost understandable to humans, yet they are not without flaws. Non-expert users, for example, may not fully understand the logic of heatmaps (the logic in which relevant pixels to the model's prediction are highlighted with different intensities or colors). Additionally, objects and regions of the input image that are relevant to the model prediction are frequently not entirely differentiated by heatmaps. In this paper, we propose a framework called TbExplain that employs XAI techniques and a pre-trained object detector to present text-based explanations of scene classification models. Moreover, TbExplain incorporates a novel method to correct predictions and textually explain them based on the statistics of objects in the input image when the initial prediction is unreliable. To assess the trustworthiness and validity of the text-based explanations, we conducted a qualitative experiment, and the findings indicated that these explanations are sufficiently reliable. Furthermore, our quantitative and qualitative experiments on TbExplain with scene classification datasets reveal an improvement in classification accuracy over ResNet variants.
Authors: Yunfei Liu, Lijian Lin, Fei Yu, Changyin Zhou, Yu Li
Audio-driven portrait animation aims to synthesize portrait videos that are conditioned by given audio. Animating high-fidelity and multimodal video portraits has a variety of applications. Previous methods have attempted to capture different motion modes and generate high-fidelity portrait videos by training different models or sampling signals from given videos. However, lacking correlation learning between lip-sync and other movements (e.g., head pose/eye blinking) usually leads to unnatural results. In this paper, we propose a unified system for multi-person, diverse, and high-fidelity talking portrait generation. Our method contains three stages, i.e., 1) Mapping-Once network with Dual Attentions (MODA) generates talking representation from given audio. In MODA, we design a dual-attention module to encode accurate mouth movements and diverse modalities. 2) Facial composer network generates dense and detailed face landmarks, and 3) temporal-guided renderer syntheses stable videos. Extensive evaluations demonstrate that the proposed system produces more natural and realistic video portraits compared to previous methods.
Authors: Ioannis Sarridis, Christos Koutlis, Symeon Papadopoulos, Christos Diou
Deep learning-based person identification and verification systems have remarkably improved in terms of accuracy in recent years; however, such systems, including widely popular cloud-based solutions, have been found to exhibit significant biases related to race, age, and gender, a problem that requires in-depth exploration and solutions. This paper presents an in-depth analysis, with a particular emphasis on the intersectionality of these demographic factors. Intersectional bias refers to the performance discrepancies w.r.t. the different combinations of race, age, and gender groups, an area relatively unexplored in current literature. Furthermore, the reliance of most state-of-the-art approaches on accuracy as the principal evaluation metric often masks significant demographic disparities in performance. To counter this crucial limitation, we incorporate five additional metrics in our quantitative analysis, including disparate impact and mistreatment metrics, which are typically ignored by the relevant fairness-aware approaches. Results on the Racial Faces in-the-Wild (RFW) benchmark indicate pervasive biases in face recognition systems, extending beyond race, with different demographic factors yielding significantly disparate outcomes. In particular, Africans demonstrate an 11.25% lower True Positive Rate (TPR) compared to Caucasians, while only a 3.51% accuracy drop is observed. Even more concerning, the intersections of multiple protected groups, such as African females over 60 years old, demonstrate a +39.89% disparate mistreatment rate compared to the highest Caucasians rate. By shedding light on these biases and their implications, this paper aims to stimulate further research towards developing fairer, more equitable face recognition and verification systems.
Authors: Jia-Xin Zhuang, Jiabin Cai, Jianguo Zhang, Wei-shi Zheng, Ruixuan Wang
Automated medical image classification is the key component in intelligent diagnosis systems. However, most medical image datasets contain plenty of samples of common diseases and just a handful of rare ones, leading to major class imbalances. Currently, it is an open problem in intelligent diagnosis to effectively learn from imbalanced training data. In this paper, we propose a simple yet effective framework, named \textbf{C}lass \textbf{A}ttention to \textbf{RE}gions of the lesion (CARE), to handle data imbalance issues by embedding attention into the training process of \textbf{C}onvolutional \textbf{N}eural \textbf{N}etworks (CNNs). The proposed attention module helps CNNs attend to lesion regions of rare diseases, therefore helping CNNs to learn their characteristics more effectively. In addition, this attention module works only during the training phase and does not change the architecture of the original network, so it can be directly combined with any existing CNN architecture. The CARE framework needs bounding boxes to represent the lesion regions of rare diseases. To alleviate the need for manual annotation, we further developed variants of CARE by leveraging the traditional saliency methods or a pretrained segmentation model for bounding box generation. Results show that the CARE variants with automated bounding box generation are comparable to the original CARE framework with \textit{manual} bounding box annotations. A series of experiments on an imbalanced skin image dataset and a pneumonia dataset indicates that our method can effectively help the network focus on the lesion regions of rare diseases and remarkably improves the classification performance of rare diseases.
Authors: Bryan Tjandra, Made S. N. Negara, Nyoo S. C. Handoko
Indonesia, as a maritime country, has a significant portion of its territory covered by water. Ineffective waste management has resulted in a considerable amount of trash in Indonesian waters, leading to various issues. The development of an automated trash-collecting robot can be a solution to address this problem. The robot requires a system capable of detecting objects in motion, such as in videos. However, using naive object detection methods in videos has limitations, particularly when image focus is reduced and the target object is obstructed by other objects. This paper's contribution provides an explanation of the methods that can be applied to perform video object detection in an automated trash-collecting robot. The study utilizes the YOLOv5 model and the Robust & Efficient Post Processing (REPP) method, along with tubelet-level bounding box linking on the FloW and Roboflow datasets. The combination of these methods enhances the performance of naive object detection from YOLOv5 by considering the detection results in adjacent frames. The results show that the post-processing stage and tubelet-level bounding box linking can improve the quality of detection, achieving approximately 3% better performance compared to YOLOv5 alone. The use of these methods has the potential to detect surface and underwater trash and can be applied to a real-time image-based trash-collecting robot. Implementing this system is expected to mitigate the damage caused by trash in the past and improve Indonesia's waste management system in the future.
Authors: Mingzhe Guo, Zhipeng Zhang, Liping Jing, Haibin Ling, Heng Fan
Multimodal vision-language (VL) learning has noticeably pushed the tendency toward generic intelligence owing to emerging large foundation models. However, tracking, as a fundamental vision problem, surprisingly enjoys less bonus from recent flourishing VL learning. We argue that the reasons are two-fold: the lack of large-scale vision-language annotated videos and ineffective vision-language interaction learning of current works. These nuisances motivate us to design more effective vision-language representation for tracking, meanwhile constructing a large database with language annotation for model learning. Particularly, in this paper, we first propose a general attribute annotation strategy to decorate videos in six popular tracking benchmarks, which contributes a large-scale vision-language tracking database with more than 23,000 videos. We then introduce a novel framework to improve tracking by learning a unified-adaptive VL representation, where the cores are the proposed asymmetric architecture search and modality mixer (ModaMixer). To further improve VL representation, we introduce a contrastive loss to align different modalities. To thoroughly evidence the effectiveness of our method, we integrate the proposed framework on three tracking methods with different designs, i.e., the CNN-based SiamCAR, the Transformer-based OSTrack, and the hybrid structure TransT. The experiments demonstrate that our framework can significantly improve all baselines on six benchmarks. Besides empirical results, we theoretically analyze our approach to show its rationality. By revealing the potential of VL representation, we expect the community to divert more attention to VL tracking and hope to open more possibilities for future tracking with diversified multimodal messages.
Authors: JoonHo Lee, Jae Oh Woo, Hankyu Moon, Kwonho Lee
Deploying deep visual models can lead to performance drops due to the discrepancies between source and target distributions. Several approaches leverage labeled source data to estimate target domain accuracy, but accessing labeled source data is often prohibitively difficult due to data confidentiality or resource limitations on serving devices. Our work proposes a new framework to estimate model accuracy on unlabeled target data without access to source data. We investigate the feasibility of using pseudo-labels for accuracy estimation and evolve this idea into adopting recent advances in source-free domain adaptation algorithms. Our approach measures the disagreement rate between the source hypothesis and the target pseudo-labeling function, adapted from the source hypothesis. We mitigate the impact of erroneous pseudo-labels that may arise due to a high ideal joint hypothesis risk by employing adaptive adversarial perturbation on the input of the target model. Our proposed source-free framework effectively addresses the challenging distribution shift scenarios and outperforms existing methods requiring source data and labels for training.
Authors: Lingting Zhu, Zeyue Xue, Zhenchao Jin, Xian Liu, Jingzhen He, Ziwei Liu, Lequan Yu
Cross-modality medical image synthesis is a critical topic and has the potential to facilitate numerous applications in the medical imaging field. Despite recent successes in deep-learning-based generative models, most current medical image synthesis methods rely on generative adversarial networks and suffer from notorious mode collapse and unstable training. Moreover, the 2D backbone-driven approaches would easily result in volumetric inconsistency, while 3D backbones are challenging and impractical due to the tremendous memory cost and training difficulty. In this paper, we introduce a new paradigm for volumetric medical data synthesis by leveraging 2D backbones and present a diffusion-based framework, Make-A-Volume, for cross-modality 3D medical image synthesis. To learn the cross-modality slice-wise mapping, we employ a latent diffusion model and learn a low-dimensional latent space, resulting in high computational efficiency. To enable the 3D image synthesis and mitigate volumetric inconsistency, we further insert a series of volumetric layers in the 2D slice-mapping model and fine-tune them with paired 3D data. This paradigm extends the 2D image diffusion model to a volumetric version with a slightly increasing number of parameters and computation, offering a principled solution for generic cross-modality 3D medical image synthesis. We showcase the effectiveness of our Make-A-Volume framework on an in-house SWI-MRA brain MRI dataset and a public T1-T2 brain MRI dataset. Experimental results demonstrate that our framework achieves superior synthesis results with volumetric consistency.
Authors: Junhao Dong, Zhu Meng, Delong Liu, Zhicheng Zhao, Fei Su
Prototype-based classification is a classical method in machine learning, and recently it has achieved remarkable success in semi-supervised semantic segmentation. However, the current approach isolates the prototype initialization process from the main training framework, which appears to be unnecessary. Furthermore, while the direct use of K-Means algorithm for prototype generation has considered rich intra-class variance, it may not be the optimal solution for the classification task. To tackle these problems, we propose a novel boundary-refined prototype generation (BRPG) method, which is incorporated into the whole training framework. Specifically, our approach samples and clusters high- and low-confidence features separately based on a confidence threshold, aiming to generate prototypes closer to the class boundaries. Moreover, an adaptive prototype optimization strategy is introduced to make prototype augmentation for categories with scattered feature distributions. Extensive experiments on the PASCAL VOC 2012 and Cityscapes datasets demonstrate the superiority and scalability of the proposed method, outperforming the current state-of-the-art approaches. The code is available at xxxxxxxxxxxxxx.
Authors: Alexandre Benatti, Luciano da F. Costa
Though performed almost effortlessly by humans, segmenting 2D gray-scale or color images in terms of their constituent regions of interest (e.g.~background, objects or portions of objects) constitutes one of the greatest challenges in science and technology as a consequence of the involved dimensionality reduction(3D to 2D), noise, reflections, shades, and occlusions, among many other possible effects. While a large number of interesting approaches have been respectively suggested along the last decades, it was mainly with the more recent development of deep learning that more effective and general solutions have been obtained, currently constituting the basic comparison reference for this type of operation. Also developed recently, a multiset-based methodology has been described that is capable of encouraging performance that combines spatial accuracy, stability, and robustness while requiring minimal computational resources (hardware and/or training and recognition time). The interesting features of the latter methodology mostly follow from the enhanced selectivity and sensitivity, as well as good robustness to data perturbations and outliers, allowed by the coincidence similarity index on which the multiset approach to supervised image segmentation is based. After describing the deep learning and multiset approaches, the present work develops two comparison experiments between them which are primarily aimed at illustrating their respective main interesting features when applied to the adopted specific type of data and parameter configurations. While the deep learning approach confirmed its potential for performing image segmentation, the alternative multiset methodology allowed for encouraging accuracy while requiring little computational resources.
Authors: Zenghao Bao, Zichang Tan, Jun Li, Jun Wan, Xibo Ma, Zhen Lei
Facial age estimation has received a lot of attention for its diverse application scenarios. Most existing studies treat each sample equally and aim to reduce the average estimation error for the entire dataset, which can be summarized as General Age Estimation. However, due to the long-tailed distribution prevalent in the dataset, treating all samples equally will inevitably bias the model toward the head classes (usually the adult with a majority of samples). Driven by this, some works suggest that each class should be treated equally to improve performance in tail classes (with a minority of samples), which can be summarized as Long-tailed Age Estimation. However, Long-tailed Age Estimation usually faces a performance trade-off, i.e., achieving improvement in tail classes by sacrificing the head classes. In this paper, our goal is to design a unified framework to perform well on both tasks, killing two birds with one stone. To this end, we propose a simple, effective, and flexible training paradigm named GLAE, which is two-fold. Our GLAE provides a surprising improvement on Morph II, reaching the lowest MAE and CMAE of 1.14 and 1.27 years, respectively. Compared to the previous best method, MAE dropped by up to 34%, which is an unprecedented improvement, and for the first time, MAE is close to 1 year old. Extensive experiments on other age benchmark datasets, including CACD, MIVIA, and Chalearn LAP 2015, also indicate that GLAE outperforms the state-of-the-art approaches significantly.
Authors: Bowen Xue, Shuang Zhao, Henrik Wann Jensen, Zahra Montazeri
Neural reflectance models are capable of accurately reproducing the spatially-varying appearance of many real-world materials at different scales. However, existing methods have difficulties handling highly glossy materials. To address this problem, we introduce a new neural reflectance model which, compared with existing methods, better preserves not only specular highlights but also fine-grained details. To this end, we enhance the neural network performance by encoding input data to frequency space, inspired by NeRF, to better preserve the details. Furthermore, we introduce a gradient-based loss and employ it in multiple stages, adaptive to the progress of the learning phase. Lastly, we utilize an optional extension to the decoder network using the Inception module for more accurate yet costly performance. We demonstrate the effectiveness of our method using a variety of synthetic and real examples.
Authors: Javad Peymanfard, Vahid Saeedi, Mohammad Reza Mohammadi, Hossein Zeinali, Nasser Mozayani
Lip reading is a challenging task that has many potential applications in speech recognition, human-computer interaction, and security systems. However, existing lip reading systems often suffer from low accuracy due to the limitations of video features. In this paper, we propose a novel approach that leverages visemes, which are groups of phonetically similar lip shapes, to extract more discriminative and robust video features for lip reading. We evaluate our approach on various tasks, including word-level and sentence-level lip reading, and audiovisual speech recognition using the Arman-AV dataset, a largescale Persian corpus. Our experimental results show that our viseme based approach consistently outperforms the state-of-theart methods in all these tasks. The proposed method reduces the lip-reading word error rate (WER) by 9.1% relative to the best previous method.
Authors: Dimitri von Rütte, Elisabetta Fedele, Jonathan Thomm, Lukas Wolf
In an era where visual content generation is increasingly driven by machine learning, the integration of human feedback into generative models presents significant opportunities for enhancing user experience and output quality. This study explores strategies for incorporating iterative human feedback into the generative process of diffusion-based text-to-image models. We propose FABRIC, a training-free approach applicable to a wide range of popular diffusion models, which exploits the self-attention layer present in the most widely used architectures to condition the diffusion process on a set of feedback images. To ensure a rigorous assessment of our approach, we introduce a comprehensive evaluation methodology, offering a robust mechanism to quantify the performance of generative visual models that integrate human feedback. We show that generation results improve over multiple rounds of iterative feedback through exhaustive analysis, implicitly optimizing arbitrary user preferences. The potential applications of these findings extend to fields such as personalized content creation and customization.
Authors: Kanghoon Lee, Jiachen Li, David Isele, Jinkyoo Park, Kikuo Fujimura, Mykel J. Kochenderfer
Although deep reinforcement learning (DRL) has shown promising results for autonomous navigation in interactive traffic scenarios, existing work typically adopts a fixed behavior policy to control social vehicles in the training environment. This may cause the learned driving policy to overfit the environment, making it difficult to interact well with vehicles with different, unseen behaviors. In this work, we introduce an efficient method to train diverse driving policies for social vehicles as a single meta-policy. By randomizing the interaction-based reward functions of social vehicles, we can generate diverse objectives and efficiently train the meta-policy through guiding policies that achieve specific objectives. We further propose a training strategy to enhance the robustness of the ego vehicle's driving policy using the environment where social vehicles are controlled by the learned meta-policy. Our method successfully learns an ego driving policy that generalizes well to unseen situations with out-of-distribution (OOD) social agents' behaviors in a challenging uncontrolled T-intersection scenario.
Authors: Moa Arvidsson, Sithichot Sawirot, Cristofer Englund, Fernando Alonso-Fernandez, Martin Torstensson, Boris Duran
Millions of vehicles are transported every year, tightly parked in vessels or boats. To reduce the risks of associated safety issues like fires, knowing the location of vehicles is essential, since different vehicles may need different mitigation measures, e.g. electric cars. This work is aimed at creating a solution based on a nano-drone that navigates across rows of parked vehicles and detects their license plates. We do so via a wall-following algorithm, and a CNN trained to detect license plates. All computations are done in real-time on the drone, which just sends position and detected images that allow the creation of a 2D map with the position of the plates. Our solution is capable of reading all plates across eight test cases (with several rows of plates, different drone speeds, or low light) by aggregation of measurements across several drone journeys.
Authors: Jiajie Fan, Laure Vuaille, Hao Wang, Thomas Bäck
Generative Engineering Design approaches driven by Deep Generative Models (DGM) have been proposed to facilitate industrial engineering processes. In such processes, designs often come in the form of images, such as blueprints, engineering drawings, and CAD models depending on the level of detail. DGMs have been successfully employed for synthesis of natural images, e.g., displaying animals, human faces and landscapes. However, industrial design images are fundamentally different from natural scenes in that they contain rich structural patterns and long-range dependencies, which are challenging for convolution-based DGMs to generate. Moreover, DGM-driven generation process is typically triggered based on random noisy inputs, which outputs unpredictable samples and thus cannot perform an efficient industrial design exploration. We tackle these challenges by proposing a novel model Self-Attention Adversarial Latent Autoencoder (SA-ALAE), which allows generating feasible design images of complex engineering parts. With SA-ALAE, users can not only explore novel variants of an existing design, but also control the generation process by operating in latent space. The potential of SA-ALAE is shown by generating engineering blueprints in a real automotive design task.
Authors: Wei Cheng, Ruixiang Chen, Wanqi Yin, Siming Fan, Keyu Chen, Honglin He, Huiwen Luo, Zhongang Cai, Jingbo Wang, Yang Gao, Zhengming Yu, Zhengyu Lin, Daxuan Ren, Lei Yang, Ziwei Liu, Chen Change Loy, Chen Qian, Wayne Wu, Dahua Lin, Bo Dai, Kwan-Yee Lin
Realistic human-centric rendering plays a key role in both computer vision and computer graphics. Rapid progress has been made in the algorithm aspect over the years, yet existing human-centric rendering datasets and benchmarks are rather impoverished in terms of diversity, which are crucial for rendering effect. Researchers are usually constrained to explore and evaluate a small set of rendering problems on current datasets, while real-world applications require methods to be robust across different scenarios. In this work, we present DNA-Rendering, a large-scale, high-fidelity repository of human performance data for neural actor rendering. DNA-Rendering presents several alluring attributes. First, our dataset contains over 1500 human subjects, 5000 motion sequences, and 67.5M frames' data volume. Second, we provide rich assets for each subject -- 2D/3D human body keypoints, foreground masks, SMPLX models, cloth/accessory materials, multi-view images, and videos. These assets boost the current method's accuracy on downstream rendering tasks. Third, we construct a professional multi-view system to capture data, which contains 60 synchronous cameras with max 4096 x 3000 resolution, 15 fps speed, and stern camera calibration steps, ensuring high-quality resources for task training and evaluation. Along with the dataset, we provide a large-scale and quantitative benchmark in full-scale, with multiple tasks to evaluate the existing progress of novel view synthesis, novel pose animation synthesis, and novel identity rendering methods. In this manuscript, we describe our DNA-Rendering effort as a revealing of new observations, challenges, and future directions to human-centric rendering. The dataset, code, and benchmarks will be publicly available at https://dna-rendering.github.io/
Authors: Xiangxiang Chu, Shun Lu, Xudong Li, Bo Zhang
Blending multiple convolutional kernels is proved advantageous in neural architecture design. However, current two-stage neural architecture search methods are mainly limited to single-path search spaces. How to efficiently search models of multi-path structures remains a difficult problem. In this paper, we are motivated to train a one-shot multi-path supernet to accurately evaluate the candidate architectures. Specifically, we discover that in the studied search spaces, feature vectors summed from multiple paths are nearly multiples of those from a single path. Such disparity perturbs the supernet training and its ranking ability. Therefore, we propose a novel mechanism called Shadow Batch Normalization (SBN) to regularize the disparate feature statistics. Extensive experiments prove that SBNs are capable of stabilizing the optimization and improving ranking performance. We call our unified multi-path one-shot approach as MixPath, which generates a series of models that achieve state-of-the-art results on ImageNet.
Authors: Yu Guo, Qiyu Jin, Gabriele Facciolo, Tieyong Zeng, Jean-Michel Morel
Image demosaicking and denoising are the first two key steps of the color image production pipeline. The classical processing sequence has for a long time consisted of applying denoising first, and then demosaicking. Applying the operations in this order leads to oversmoothing and checkerboard effects. Yet, it was difficult to change this order, because once the image is demosaicked, the statistical properties of the noise are dramatically changed and hard to handle by traditional denoising models. In this paper, we address this problem by a hybrid machine learning method. We invert the traditional color filter array (CFA) processing pipeline by first demosaicking and then denoising. Our demosaicking algorithm, trained on noiseless images, combines a traditional method and a residual convolutional neural network (CNN). This first stage retains all known information, which is the key point to obtain faithful final results. The noisy demosaicked image is then passed through a second CNN restoring a noiseless full-color image. This pipeline order completely avoids checkerboard effects and restores fine image detail. Although CNNs can be trained to solve jointly demosaicking-denoising end-to-end, we find that this two-stage training performs better and is less prone to failure. It is shown experimentally to improve on the state of the art, both quantitatively and in terms of visual quality.
Authors: Aleksandar Vakanski, Min Xian
The generalization performance of deep learning models for medical image analysis often decreases on images collected with different devices for data acquisition, device settings, or patient population. A better understanding of the generalization capacity on new images is crucial for clinicians' trustworthiness in deep learning. Although significant research efforts have been recently directed toward establishing generalization bounds and complexity measures, still, there is often a significant discrepancy between the predicted and actual generalization performance. As well, related large empirical studies have been primarily based on validation with general-purpose image datasets. This paper presents an empirical study that investigates the correlation between 25 complexity measures and the generalization abilities of supervised deep learning classifiers for breast ultrasound images. The results indicate that PAC-Bayes flatness-based and path norm-based measures produce the most consistent explanation for the combination of models and data. We also investigate the use of multi-task classification and segmentation approach for breast images, and report that such learning approach acts as an implicit regularizer and is conducive toward improved generalization.
This paper concerns the research problem of point cloud registration to find the rigid transformation to optimally align the source point set with the target one. Learning robust point cloud registration models with deep neural networks has emerged as a powerful paradigm, offering promising performance in predicting the global geometric transformation for a pair of point sets. Existing methods firstly leverage an encoder to regress a latent shape embedding, which is then decoded into a shape-conditioned transformation via concatenation-based conditioning. However, different regions of a 3D shape vary in their geometric structures which makes it more sense that we have a region-conditioned transformation instead of the shape-conditioned one. In this paper we present a \underline{R}egion-\underline{A}ware point cloud \underline{R}egistration, denoted as RAR, to predict transformation for pairwise point sets in the self-supervised learning fashion. More specifically, we develop a novel region-aware decoder (RAD) module that is formed with an implicit neural region representation parameterized by neural networks. The implicit neural region representation is learned with a self-supervised 3D shape reconstruction loss without the need for region labels. Consequently, the region-aware decoder (RAD) module guides the training of the region-aware transformation (RAT) module and region-aware weight (RAW) module, which predict the transforms and weights for different regions respectively. The global geometric transformation from source point set to target one is then formed by the weighted fusion of region-aware transforms. Compared to the state-of-the-art approaches, our experiments show that our RAR achieves superior registration performance over various benchmark datasets (e.g. ModelNet40).
Authors: Hao Su, Jianwei Niu, Xuefeng Liu, Jiahe Cui, Ji Wan
Manga is a fashionable Japanese-style comic form that is composed of black-and-white strokes and is generally displayed as raster images on digital devices. Typical mangas have simple textures, wide lines, and few color gradients, which are vectorizable natures to enjoy the merits of vector graphics, e.g., adaptive resolutions and small file sizes. In this paper, we propose MARVEL (MAnga's Raster to VEctor Learning), a primitive-wise approach for vectorizing raster mangas by Deep Reinforcement Learning (DRL). Unlike previous learning-based methods which predict vector parameters for an entire image, MARVEL introduces a new perspective that regards an entire manga as a collection of basic primitives\textemdash stroke lines, and designs a DRL model to decompose the target image into a primitive sequence for achieving accurate vectorization. To improve vectorization accuracies and decrease file sizes, we further propose a stroke accuracy reward to predict accurate stroke lines, and a pruning mechanism to avoid generating erroneous and repeated strokes. Extensive subjective and objective experiments show that our MARVEL can generate impressive results and reaches the state-of-the-art level. Our code is open-source at: https://github.com/SwordHolderSH/Mang2Vec.
Authors: Peng Zhi, Haoran Zhou, Hang Huang, Rui Zhao, Rui Zhou, Qingguo Zhou
In the field of state-of-the-art object detection, the task of object localization is typically accomplished through a dedicated subnet that emphasizes bounding box regression. This subnet traditionally predicts the object's position by regressing the box's center position and scaling factors. Despite the widespread adoption of this approach, we have observed that the localization results often suffer from defects, leading to unsatisfactory detector performance. In this paper, we address the shortcomings of previous methods through theoretical analysis and experimental verification and present an innovative solution for precise object detection. Instead of solely focusing on the object's center and size, our approach enhances the accuracy of bounding box localization by refining the box edges based on the estimated distribution at the object's boundary. Experimental results demonstrate the potential and generalizability of our proposed method.
Authors: Michael P. J. Camilleri, Li Zhang, Rasneer S. Bains, Andrew Zisserman, Christopher K. I. Williams
Our objective is to locate and provide a unique identifier for each mouse in a cluttered home-cage environment through time, as a precursor to automated behaviour recognition for biological research. This is a very challenging problem due to (i) the lack of distinguishing visual features for each mouse, and (ii) the close confines of the scene with constant occlusion, making standard visual tracking approaches unusable. However, a coarse estimate of each mouse's location is available from a unique RFID implant, so there is the potential to optimally combine information from (weak) tracking with coarse information on identity. To achieve our objective, we make the following key contributions: (a) the formulation of the object identification problem as an assignment problem (solved using Integer Linear Programming), and (b) a novel probabilistic model of the affinity between tracklets and RFID data. The latter is a crucial part of the model, as it provides a principled probabilistic treatment of object detections given coarse localisation. Our approach achieves 77% accuracy on this animal identification problem, and is able to reject spurious detections when the animals are hidden.
Authors: Haekyu Park, Seongmin Lee, Benjamin Hoover, Austin Wright, Omar Shaikh, Rahul Duggal, Nilaksh Das, Judy Hoffman, Duen Horng Chau
We present ConceptEvo, a unified interpretation framework for deep neural networks (DNNs) that reveals the inception and evolution of learned concepts during training. Our work fills a critical gap in DNN interpretation research, as existing methods focus on post-hoc interpretation after training. ConceptEvo presents two novel technical contributions: (1) an algorithm that generates a unified semantic space that enables side-by-side comparison of different models during training; and (2) an algorithm that discovers and quantifies important concept evolutions for class predictions. Through a large-scale human evaluation with 260 participants and quantitative experiments, we show that ConceptEvo discovers evolutions across different models that are meaningful to humans and important for predictions. ConceptEvo works for both modern (ConvNeXt) and classic DNNs (e.g., VGGs, InceptionV3).
Authors: Mingbao Lin, Mengzhao Chen, Yuxin Zhang, Chunhua Shen, Rongrong Ji, Liujuan Cao
We attempt to reduce the computational costs in vision transformers (ViTs), which increase quadratically in the token number. We present a novel training paradigm that trains only one ViT model at a time, but is capable of providing improved image recognition performance with various computational costs. Here, the trained ViT model, termed super vision transformer (SuperViT), is empowered with the versatile ability to solve incoming patches of multiple sizes as well as preserve informative tokens with multiple keeping rates (the ratio of keeping tokens) to achieve good hardware efficiency for inference, given that the available hardware resources often change from time to time. Experimental results on ImageNet demonstrate that our SuperViT can considerably reduce the computational costs of ViT models with even performance increase. For example, we reduce 2x FLOPs of DeiT-S while increasing the Top-1 accuracy by 0.2% and 0.7% for 1.5x reduction. Also, our SuperViT significantly outperforms existing studies on efficient vision transformers. For example, when consuming the same amount of FLOPs, our SuperViT surpasses the recent state-of-the-art (SOTA) EViT by 1.1% when using DeiT-S as their backbones. The project of this work is made publicly available at https://github.com/lmbxmu/SuperViT.
Authors: Madeline C. Schiappa, Yogesh S. Rawat, Mubarak Shah
The remarkable success of deep learning in various domains relies on the availability of large-scale annotated datasets. However, obtaining annotations is expensive and requires great effort, which is especially challenging for videos. Moreover, the use of human-generated annotations leads to models with biased learning and poor domain generalization and robustness. As an alternative, self-supervised learning provides a way for representation learning which does not require annotations and has shown promise in both image and video domains. Different from the image domain, learning video representations are more challenging due to the temporal dimension, bringing in motion and other environmental dynamics. This also provides opportunities for video-exclusive ideas that advance self-supervised learning in the video and multimodal domain. In this survey, we provide a review of existing approaches on self-supervised learning focusing on the video domain. We summarize these methods into four different categories based on their learning objectives: 1) pretext tasks, 2) generative learning, 3) contrastive learning, and 4) cross-modal agreement. We further introduce the commonly used datasets, downstream evaluation tasks, insights into the limitations of existing works, and the potential future directions in this area.
Authors: Jungho Lee, Minhyeok Lee, Dogyoon Lee, Sangyoun Lee
Graph convolutional networks (GCNs) are the most commonly used methods for skeleton-based action recognition and have achieved remarkable performance. Generating adjacency matrices with semantically meaningful edges is particularly important for this task, but extracting such edges is challenging problem. To solve this, we propose a hierarchically decomposed graph convolutional network (HD-GCN) architecture with a novel hierarchically decomposed graph (HD-Graph). The proposed HD-GCN effectively decomposes every joint node into several sets to extract major structurally adjacent and distant edges, and uses them to construct an HD-Graph containing those edges in the same semantic spaces of a human skeleton. In addition, we introduce an attention-guided hierarchy aggregation (A-HA) module to highlight the dominant hierarchical edge sets of the HD-Graph. Furthermore, we apply a new six-way ensemble method, which uses only joint and bone stream without any motion stream. The proposed model is evaluated and achieves state-of-the-art performance on four large, popular datasets. Finally, we demonstrate the effectiveness of our model with various comparative experiments.
Authors: Ashwin De Silva, Rahul Ramesh, Carey E. Priebe, Pratik Chaudhari, Joshua T. Vogelstein
We expect the generalization error to improve with more samples from a similar task, and to deteriorate with more samples from an out-of-distribution (OOD) task. In this work, we show a counter-intuitive phenomenon: the generalization error of a task can be a non-monotonic function of the number of OOD samples. As the number of OOD samples increases, the generalization error on the target task improves before deteriorating beyond a threshold. In other words, there is value in training on small amounts of OOD data. We use Fisher's Linear Discriminant on synthetic datasets and deep networks on computer vision benchmarks such as MNIST, CIFAR-10, CINIC-10, PACS and DomainNet to demonstrate and analyze this phenomenon. In the idealistic setting where we know which samples are OOD, we show that these non-monotonic trends can be exploited using an appropriately weighted objective of the target and OOD empirical risk. While its practical utility is limited, this does suggest that if we can detect OOD samples, then there may be ways to benefit from them. When we do not know which samples are OOD, we show how a number of go-to strategies such as data-augmentation, hyper-parameter optimization, and pre-training are not enough to ensure that the target generalization error does not deteriorate with the number of OOD samples in the dataset.
Authors: Chengyin Li, Yao Qiang, Rafi Ibn Sultan, Hassan Bagher-Ebadian, Prashant Khanduri, Indrin J. Chetty, Dongxiao Zhu
Computed Tomography (CT) based precise prostate segmentation for treatment planning is challenging due to (1) the unclear boundary of the prostate derived from CT's poor soft tissue contrast and (2) the limitation of convolutional neural network-based models in capturing long-range global context. Here we propose a novel focal transformer-based image segmentation architecture to effectively and efficiently extract local visual features and global context from CT images. Additionally, we design an auxiliary boundary-induced label regression task coupled with the main prostate segmentation task to address the unclear boundary issue in CT images. We demonstrate that this design significantly improves the quality of the CT-based prostate segmentation task over other competing methods, resulting in substantially improved performance, i.e., higher Dice Similarity Coefficient, lower Hausdorff Distance, and Average Symmetric Surface Distance, on both private and public CT image datasets. Our code is available at this \href{https://github.com/ChengyinLee/FocalUNETR.git}{link}.
Authors: Wentao Zhu, Xiaoxuan Ma, Zhaoyang Liu, Libin Liu, Wayne Wu, Yizhou Wang
We present a unified perspective on tackling various human-centric video tasks by learning human motion representations from large-scale and heterogeneous data resources. Specifically, we propose a pretraining stage in which a motion encoder is trained to recover the underlying 3D motion from noisy partial 2D observations. The motion representations acquired in this way incorporate geometric, kinematic, and physical knowledge about human motion, which can be easily transferred to multiple downstream tasks. We implement the motion encoder with a Dual-stream Spatio-temporal Transformer (DSTformer) neural network. It could capture long-range spatio-temporal relationships among the skeletal joints comprehensively and adaptively, exemplified by the lowest 3D pose estimation error so far when trained from scratch. Furthermore, our proposed framework achieves state-of-the-art performance on all three downstream tasks by simply finetuning the pretrained motion encoder with a simple regression head (1-2 layers), which demonstrates the versatility of the learned motion representations. Code and models are available at https://motionbert.github.io/
Authors: Marc Habermann, Lingjie Liu, Weipeng Xu, Gerard Pons-Moll, Michael Zollhoefer, Christian Theobalt
Photo-real digital human avatars are of enormous importance in graphics, as they enable immersive communication over the globe, improve gaming and entertainment experiences, and can be particularly beneficial for AR and VR settings. However, current avatar generation approaches either fall short in high-fidelity novel view synthesis, generalization to novel motions, reproduction of loose clothing, or they cannot render characters at the high resolution offered by modern displays. To this end, we propose HDHumans, which is the first method for HD human character synthesis that jointly produces an accurate and temporally coherent 3D deforming surface and highly photo-realistic images of arbitrary novel views and of motions not seen at training time. At the technical core, our method tightly integrates a classical deforming character template with neural radiance fields (NeRF). Our method is carefully designed to achieve a synergy between classical surface deformation and NeRF. First, the template guides the NeRF, which allows synthesizing novel views of a highly dynamic and articulated character and even enables the synthesis of novel motions. Second, we also leverage the dense pointclouds resulting from NeRF to further improve the deforming surface via 3D-to-3D supervision. We outperform the state of the art quantitatively and qualitatively in terms of synthesis quality and resolution, as well as the quality of 3D surface reconstruction.
Authors: Fengfan Zhou, Hefei Ling, Yuxuan Shi, Jiazhong Chen, Zongyi Li, Ping Li
Face recognition (FR) models can be easily fooled by adversarial examples, which are crafted by adding imperceptible perturbations on benign face images. The existence of adversarial face examples poses a great threat to the security of society. In order to build a more sustainable digital nation, in this paper, we improve the transferability of adversarial face examples to expose more blind spots of existing FR models. Though generating hard samples has shown its effectiveness in improving the generalization of models in training tasks, the effectiveness of utilizing this idea to improve the transferability of adversarial face examples remains unexplored. To this end, based on the property of hard samples and the symmetry between training tasks and adversarial attack tasks, we propose the concept of hard models, which have similar effects as hard samples for adversarial attack tasks. Utilizing the concept of hard models, we propose a novel attack method called Beneficial Perturbation Feature Augmentation Attack (BPFA), which reduces the overfitting of adversarial examples to surrogate FR models by constantly generating new hard models to craft the adversarial examples. Specifically, in the backpropagation, BPFA records the gradients on pre-selected feature maps and uses the gradient on the input image to craft the adversarial example. In the next forward propagation, BPFA leverages the recorded gradients to add beneficial perturbations on their corresponding feature maps to increase the loss. Extensive experiments demonstrate that BPFA can significantly boost the transferability of adversarial attacks on FR.
Authors: Jungho Lee, Minhyeok Lee, Suhwan Cho, Sungmin Woo, Sungjun Jang, Sangyoun Lee
Skeleton-based action recognition has attracted considerable attention due to its compact representation of the human body's skeletal sructure. Many recent methods have achieved remarkable performance using graph convolutional networks (GCNs) and convolutional neural networks (CNNs), which extract spatial and temporal features, respectively. Although spatial and temporal dependencies in the human skeleton have been explored separately, spatio-temporal dependency is rarely considered. In this paper, we propose the Spatio-Temporal Curve Network (STC-Net) to effectively leverage the spatio-temporal dependency of the human skeleton. Our proposed network consists of two novel elements: 1) The Spatio-Temporal Curve (STC) module; and 2) Dilated Kernels for Graph Convolution (DK-GC). The STC module dynamically adjusts the receptive field by identifying meaningful node connections between every adjacent frame and generating spatio-temporal curves based on the identified node connections, providing an adaptive spatio-temporal coverage. In addition, we propose DK-GC to consider long-range dependencies, which results in a large receptive field without any additional parameters by applying an extended kernel to the given adjacency matrices of the graph. Our STC-Net combines these two modules and achieves state-of-the-art performance on four skeleton-based action recognition benchmarks.
Authors: Kumar Ashutosh, Rohit Girdhar, Lorenzo Torresani, Kristen Grauman
Narrated ''how-to'' videos have emerged as a promising data source for a wide range of learning problems, from learning visual representations to training robot policies. However, this data is extremely noisy, as the narrations do not always describe the actions demonstrated in the video. To address this problem we introduce the novel task of visual narration detection, which entails determining whether a narration is visually depicted by the actions in the video. We propose What You Say is What You Show (WYS^2), a method that leverages multi-modal cues and pseudo-labeling to learn to detect visual narrations with only weakly labeled data. Our model successfully detects visual narrations in in-the-wild videos, outperforming strong baselines, and we demonstrate its impact for state-of-the-art summarization and temporal alignment of instructional videos.
Authors: Xia Xu, Jochen Triesch
Self-supervised representation learning (SSRL) methods have shown great success in computer vision. In recent studies, augmentation-based contrastive learning methods have been proposed for learning representations that are invariant or equivariant to pre-defined data augmentation operations. However, invariant or equivariant features favor only specific downstream tasks depending on the augmentations chosen. They may result in poor performance when the learned representation does not match task requirements. Here, we consider an active observer that can manipulate views of an object and has knowledge of the action(s) that generated each view. We introduce Contrastive Invariant and Predictive Equivariant Representation learning (CIPER). CIPER comprises both invariant and equivariant learning objectives using one shared encoder and two different output heads on top of the encoder. One output head is a projection head with a state-of-the-art contrastive objective to encourage invariance to augmentations. The other is a prediction head estimating the augmentation parameters, capturing equivariant features. Both heads are discarded after training and only the encoder is used for downstream tasks. We evaluate our method on static image tasks and time-augmented image datasets. Our results show that CIPER outperforms a baseline contrastive method on various tasks. Interestingly, CIPER encourages the formation of hierarchically structured representations where different views of an object become systematically organized in the latent representation space.
Authors: Qizhang Li, Yiwen Guo, Wangmeng Zuo, Hao Chen
The transferability of adversarial examples across deep neural networks (DNNs) is the crux of many black-box attacks. Many prior efforts have been devoted to improving the transferability via increasing the diversity in inputs of some substitute models. In this paper, by contrast, we opt for the diversity in substitute models and advocate to attack a Bayesian model for achieving desirable transferability. Deriving from the Bayesian formulation, we develop a principled strategy for possible finetuning, which can be combined with many off-the-shelf Gaussian posterior approximations over DNN parameters. Extensive experiments have been conducted to verify the effectiveness of our method, on common benchmark datasets, and the results demonstrate that our method outperforms recent state-of-the-arts by large margins (roughly 19% absolute increase in average attack success rate on ImageNet), and, by combining with these recent methods, further performance gain can be obtained. Our code: https://github.com/qizhangli/MoreBayesian-attack.
Authors: Xu Zheng, Yexin Liu, Yunfan Lu, Tongyan Hua, Tianbo Pan, Weiming Zhang, Dacheng Tao, Lin Wang
Event cameras are bio-inspired sensors that capture the per-pixel intensity changes asynchronously and produce event streams encoding the time, pixel position, and polarity (sign) of the intensity changes. Event cameras possess a myriad of advantages over canonical frame-based cameras, such as high temporal resolution, high dynamic range, low latency, etc. Being capable of capturing information in challenging visual conditions, event cameras have the potential to overcome the limitations of frame-based cameras in the computer vision and robotics community. In very recent years, deep learning (DL) has been brought to this emerging field and inspired active research endeavors in mining its potential. However, there is still a lack of taxonomies in DL techniques for event-based vision. We first scrutinize the typical event representations with quality enhancement methods as they play a pivotal role as inputs to the DL models. We then provide a comprehensive taxonomy for existing DL-based methods by structurally grouping them into two major categories: 1) image reconstruction and restoration; 2) event-based scene understanding and 3D vision. Importantly, we conduct benchmark experiments for the existing methods in some representative research directions (eg, object recognition) to identify some critical insights and problems. Finally, we make important discussions regarding the challenges and provide new perspectives for inspiring more research studies.
Authors: Haofei Zhang, Mengqi Xue, Xiaokang Liu, Kaixuan Chen, Jie Song, Mingli Song
In this paper, we study a novel inference paradigm, termed as schema inference, that learns to deductively infer the explainable predictions by rebuilding the prior deep neural network (DNN) forwarding scheme, guided by the prevalent philosophical cognitive concept of schema. We strive to reformulate the conventional model inference pipeline into a graph matching policy that associates the extracted visual concepts of an image with the pre-computed scene impression, by analogy with human reasoning mechanism via impression matching. To this end, we devise an elaborated architecture, termed as SchemaNet, as a dedicated instantiation of the proposed schema inference concept, that models both the visual semantics of input instances and the learned abstract imaginations of target categories as topological relational graphs. Meanwhile, to capture and leverage the compositional contributions of visual semantics in a global view, we also introduce a universal Feat2Graph scheme in SchemaNet to establish the relational graphs that contain abundant interaction information. Both the theoretical analysis and the experimental results on several benchmarks demonstrate that the proposed schema inference achieves encouraging performance and meanwhile yields a clear picture of the deductive process leading to the predictions. Our code is available at https://github.com/zhfeing/SchemaNet-PyTorch.
Authors: Axel Levy, Mark Matthews, Matan Sela, Gordon Wetzstein, Dmitry Lagun
Neural radiance fields enable novel-view synthesis and scene reconstruction with photorealistic quality from a few images, but require known and accurate camera poses. Conventional pose estimation algorithms fail on smooth or self-similar scenes, while methods performing inverse rendering from unposed views require a rough initialization of the camera orientations. The main difficulty of pose estimation lies in real-life objects being almost invariant under certain transformations, making the photometric distance between rendered views non-convex with respect to the camera parameters. Using an equivalence relation that matches the distribution of local minima in camera space, we reduce this space to its quotient set, in which pose estimation becomes a more convex problem. Using a neural-network to regularize pose estimation, we demonstrate that our method - MELON - can reconstruct a neural radiance field from unposed images with state-of-the-art accuracy while requiring ten times fewer views than adversarial approaches.
Authors: Johannes Thalhammer, Manuel Schultheiss, Tina Dorosti, Tobias Lasser, Franz Pfeiffer, Daniela Pfeiffer, Florian Schaff
Purpose: Sparse-view computed tomography (CT) is an effective way to reduce dose by lowering the total number of views acquired, albeit at the expense of image quality, which, in turn, can impact the ability to detect diseases. We explore deep learning-based artifact reduction in sparse-view cranial CT scans and its impact on automated hemorrhage detection. Methods: We trained a U-Net for artefact reduction on simulated sparse-view cranial CT scans from 3000 patients obtained from a public dataset and reconstructed with varying levels of sub-sampling. Additionally, we trained a convolutional neural network on fully sampled CT data from 17,545 patients for automated hemorrhage detection. We evaluated the classification performance using the area under the receiver operator characteristic curves (AUC-ROCs) with corresponding 95% confidence intervals (CIs) and the DeLong test, along with confusion matrices. The performance of the U-Net was compared to an analytical approach based on total variation (TV). Results: The U-Net performed superior compared to unprocessed and TV-processed images with respect to image quality and automated hemorrhage diagnosis. With U-Net post-processing, the number of views can be reduced from 4096 (AUC-ROC: 0.974; 95% CI: 0.972-0.976) views to 512 views (0.973; 0.971-0.975) with minimal decrease in hemorrhage detection (P<.001) and to 256 views (0.967; 0.964-0.969) with a slight performance decrease (P<.001). Conclusion: The results suggest that U-Net based artifact reduction substantially enhances automated hemorrhage detection in sparse-view cranial CTs. Our findings highlight that appropriate post-processing is crucial for optimal image quality and diagnostic accuracy while minimizing radiation dose.
Authors: Jing Zhao, Heliang Zheng, Chaoyue Wang, Long Lan, Wenjing Yang
The advent of open-source AI communities has produced a cornucopia of powerful text-guided diffusion models that are trained on various datasets. While few explorations have been conducted on ensembling such models to combine their strengths. In this work, we propose a simple yet effective method called Saliency-aware Noise Blending (SNB) that can empower the fused text-guided diffusion models to achieve more controllable generation. Specifically, we experimentally find that the responses of classifier-free guidance are highly related to the saliency of generated images. Thus we propose to trust different models in their areas of expertise by blending the predicted noises of two diffusion models in a saliency-aware manner. SNB is training-free and can be completed within a DDIM sampling process. Additionally, it can automatically align the semantics of two noise spaces without requiring additional annotations such as masks. Extensive experiments show the impressive effectiveness of SNB in various applications. Project page is available at https://magicfusion.github.io/.
Authors: Jianhui Liu, Yukang Chen, Xiaoqing Ye, Xiaojuan Qi
Category-level 6D pose estimation aims to predict the poses and sizes of unseen objects from a specific category. Thanks to prior deformation, which explicitly adapts a category-specific 3D prior (i.e., a 3D template) to a given object instance, prior-based methods attained great success and have become a major research stream. However, obtaining category-specific priors requires collecting a large amount of 3D models, which is labor-consuming and often not accessible in practice. This motivates us to investigate whether priors are necessary to make prior-based methods effective. Our empirical study shows that the 3D prior itself is not the credit to the high performance. The keypoint actually is the explicit deformation process, which aligns camera and world coordinates supervised by world-space 3D models (also called canonical space). Inspired by these observations, we introduce a simple prior-free implicit space transformation network, namely IST-Net, to transform camera-space features to world-space counterparts and build correspondence between them in an implicit manner without relying on 3D priors. Besides, we design camera- and world-space enhancers to enrich the features with pose-sensitive information and geometrical constraints, respectively. Albeit simple, IST-Net achieves state-of-the-art performance based-on prior-free design, with top inference speed on the REAL275 benchmark. Our code and models are available at https://github.com/CVMI-Lab/IST-Net.
Authors: Luigi Riz, Andrea Caraffa, Matteo Bortolon, Mohamed Lamine Mekhalfi, Davide Boscaini, André Moura, José Antunes, André Dias, Hugo Silva, Andreas Leonidou, Christos Constantinides, Christos Keleshis, Dante Abate, Fabio Poiesi
We present MONET, a new multimodal dataset captured using a thermal camera mounted on a drone that flew over rural areas, and recorded human and vehicle activities. We captured MONET to study the problem of object localisation and behaviour understanding of targets undergoing large-scale variations and being recorded from different and moving viewpoints. Target activities occur in two different land sites, each with unique scene structures and cluttered backgrounds. MONET consists of approximately 53K images featuring 162K manually annotated bounding boxes. Each image is timestamp-aligned with drone metadata that includes information about attitudes, speed, altitude, and GPS coordinates. MONET is different from previous thermal drone datasets because it features multimodal data, including rural scenes captured with thermal cameras containing both person and vehicle targets, along with trajectory information and metadata. We assessed the difficulty of the dataset in terms of transfer learning between the two sites and evaluated nine object detection algorithms to identify the open challenges associated with this type of data. Project page: https://github.com/fabiopoiesi/monet_dataset.
Authors: E. Bueno-Benito, B. Tura, M. Dimiccoli
In this paper, we propose a novel fully unsupervised framework that learns action representations suitable for the action segmentation task from the single input video itself, without requiring any training data. Our method is a deep metric learning approach rooted in a shallow network with a triplet loss operating on similarity distributions and a novel triplet selection strategy that effectively models temporal and semantic priors to discover actions in the new representational space. Under these circumstances, we successfully recover temporal boundaries in the learned action representations with higher quality compared with existing unsupervised approaches. The proposed method is evaluated on two widely used benchmark datasets for the action segmentation task and it achieves competitive performance by applying a generic clustering algorithm on the learned representations.
Authors: Seulki Park, Daeho Um, Hajung Yoon, Sanghyuk Chun, Sangdoo Yun, Jin Young Choi
In this paper, we propose a robustness benchmark for image-text matching models to assess their vulnerabilities. To this end, we insert adversarial texts and images into the search pool (i.e., gallery set) and evaluate models with the adversarial data. Specifically, we replace a word in the text to change the meaning of the text and mix images with different images to create perceptible changes in pixels. We assume that such explicit alterations would not deceive a robust model, as they should understand the holistic meaning of texts and images simultaneously. However, in our evaluations on the proposed benchmark, many state-of-the-art models show significant performance degradation, e.g., Recall@1: 81.9% $\rightarrow$ 64.5% in BLIP, 66.1% $\rightarrow$ 37.5% in VSE$\infty$, where the models favor adversarial texts/images over the original ones. This reveals the current vision-language models may not account for subtle changes or understand the overall context of texts and images. Our findings can provide insights for improving the robustness of the vision-language models and devising more diverse stress-test methods in cross-modal retrieval task. Source code and dataset will be available at https://github.com/pseulki/rococo.
Authors: H M Dipu Kabir
Multitask learning is a popular approach to training high-performing neural networks with improved generalization. In this paper, we propose a background class to achieve improved generalization at a lower computation compared to multitask learning to help researchers and organizations with limited computation power. We also present a methodology for selecting background images and discuss potential future improvements. We apply our approach to several datasets and achieved improved generalization with much lower computation. We also investigate class activation mappings (CAMs) of the trained model and observed the tendency towards looking at a bigger picture in a few class classification problems with the proposed model training methodology. Applying transformer with the proposed background class, we receive state-of-the-art (SOTA) performance on STL-10, Caltech-101, and CINIC-10 datasets. Example scripts are available in the `CAM' folder of the following GitHub Repository: github.com/dipuk0506/UQ
Authors: Momina Liaqat Ali, Zunaira Rauf, Asifullah Khan, Anabia Sohail, Rafi Ullah, Jeonghwan Gwak
Transformers, due to their ability to learn long range dependencies, have overcome the shortcomings of convolutional neural networks (CNNs) for global perspective learning. Therefore, they have gained the focus of researchers for several vision related tasks including medical diagnosis. However, their multi-head attention module only captures global level feature representations, which is insufficient for medical images. To address this issue, we propose a Channel Boosted Hybrid Vision Transformer (CB HVT) that uses transfer learning to generate boosted channels and employs both transformers and CNNs to analyse lymphocytes in histopathological images. The proposed CB HVT comprises five modules, including a channel generation module, channel exploitation module, channel merging module, region-aware module, and a detection and segmentation head, which work together to effectively identify lymphocytes. The channel generation module uses the idea of channel boosting through transfer learning to extract diverse channels from different auxiliary learners. In the CB HVT, these boosted channels are first concatenated and ranked using an attention mechanism in the channel exploitation module. A fusion block is then utilized in the channel merging module for a gradual and systematic merging of the diverse boosted channels to improve the network's learning representations. The CB HVT also employs a proposal network in its region aware module and a head to effectively identify objects, even in overlapping regions and with artifacts. We evaluated the proposed CB HVT on two publicly available datasets for lymphocyte assessment in histopathological images. The results show that CB HVT outperformed other state of the art detection models, and has good generalization ability, demonstrating its value as a tool for pathologists.
Authors: Mingyuan Meng, Bingxin Gu, Michael Fulham, Shaoli Song, Dagan Feng, Lei Bi, Jinman Kim
Survival prediction is a major concern for cancer management. Deep survival models based on deep learning have been widely adopted to perform end-to-end survival prediction from medical images. Recent deep survival models achieved promising performance by jointly performing tumor segmentation with survival prediction, where the models were guided to extract tumor-related information through Multi-Task Learning (MTL). However, these deep survival models have difficulties in exploring out-of-tumor prognostic information. In addition, existing deep survival models are unable to effectively leverage multi-modality images. Empirically-designed fusion strategies were commonly adopted to fuse multi-modality information via task-specific manually-designed networks, thus limiting the adaptability to different scenarios. In this study, we propose an Adaptive Multi-modality Segmentation-to-Survival model (AdaMSS) for survival prediction from PET/CT images. Instead of adopting MTL, we propose a novel Segmentation-to-Survival Learning (SSL) strategy, where our AdaMSS is trained for tumor segmentation and survival prediction sequentially in two stages. This strategy enables the AdaMSS to focus on tumor regions in the first stage and gradually expand its focus to include other prognosis-related regions in the second stage. We also propose a data-driven strategy to fuse multi-modality information, which realizes adaptive optimization of fusion strategies based on training data during training. With the SSL and data-driven fusion strategies, our AdaMSS is designed as an adaptive model that can self-adapt its focus regions and fusion strategy for different training stages. Extensive experiments with two large clinical datasets show that our AdaMSS outperforms state-of-the-art survival prediction methods.
Authors: Aihua Zheng, Zhiqi Ma, Zi Wang, Chenglong Li
Multi-spectral vehicle re-identification aims to address the challenge of identifying vehicles in complex lighting conditions by incorporating complementary visible and infrared information. However, in harsh environments, the discriminative cues in RGB and NIR modalities are often lost due to strong flares from vehicle lamps or sunlight, and existing multi-modal fusion methods are limited in their ability to recover these important cues. To address this problem, we propose a Flare-Aware Cross-modal Enhancement Network that adaptively restores flare-corrupted RGB and NIR features with guidance from the flare-immunized thermal infrared spectrum. First, to reduce the influence of locally degraded appearance due to intense flare, we propose a Mutual Flare Mask Prediction module to jointly obtain flare-corrupted masks in RGB and NIR modalities in a self-supervised manner. Second, to use the flare-immunized TI information to enhance the masked RGB and NIR, we propose a Flare-Aware Cross-modal Enhancement module that adaptively guides feature extraction of masked RGB and NIR spectra with prior flare-immunized knowledge from the TI spectrum. Third, to extract common informative semantic information from RGB and NIR, we propose an Inter-modality Consistency loss that enforces semantic consistency between the two modalities. Finally, to evaluate the proposed FACENet in handling intense flare, we introduce a new multi-spectral vehicle re-ID dataset, called WMVEID863, with additional challenges such as motion blur, significant background changes, and particularly intense flare degradation. Comprehensive experiments on both the newly collected dataset and public benchmark multi-spectral vehicle re-ID datasets demonstrate the superior performance of the proposed FACENet compared to state-of-the-art methods, especially in handling strong flares. The code and dataset will be released at this link.
Authors: Haojun Yu, Youcheng Li, QuanLin Wu, Ziwei Zhao, Dengbo Chen, Dong Wang, Liwei Wang
During ultrasonic scanning processes, real-time lesion detection can assist radiologists in accurate cancer diagnosis. However, this essential task remains challenging and underexplored. General-purpose real-time object detection models can mistakenly report obvious false positives (FPs) when applied to ultrasound videos, potentially misleading junior radiologists. One key issue is their failure to utilize negative symptoms in previous frames, denoted as negative temporal contexts (NTC). To address this issue, we propose to extract contexts from previous frames, including NTC, with the guidance of inverse optical flow. By aggregating extracted contexts, we endow the model with the ability to suppress FPs by leveraging NTC. We call the resulting model UltraDet. The proposed UltraDet demonstrates significant improvement over previous state-of-the-arts and achieves real-time inference speed. We release the code, checkpoints, and high-quality labels of the CVA-BUS dataset in https://github.com/HaojunYu1998/UltraDet.
Authors: Raz Lapid, Moshe Sipper
Modern image-to-text systems typically adopt the encoder-decoder framework, which comprises two main components: an image encoder, responsible for extracting image features, and a transformer-based decoder, used for generating captions. Taking inspiration from the analysis of neural networks' robustness against adversarial perturbations, we propose a novel gray-box algorithm for creating adversarial examples in image-to-text models. Unlike image classification tasks that have a finite set of class labels, finding visually similar adversarial examples in an image-to-text task poses greater challenges because the captioning system allows for a virtually infinite space of possible captions. In this paper, we present a gray-box adversarial attack on image-to-text, both untargeted and targeted. We formulate the process of discovering adversarial perturbations as an optimization problem that uses only the image-encoder component, meaning the proposed attack is language-model agnostic. Through experiments conducted on the ViT-GPT2 model, which is the most-used image-to-text model in Hugging Face, and the Flickr30k dataset, we demonstrate that our proposed attack successfully generates visually similar adversarial examples, both with untargeted and targeted captions. Notably, our attack operates in a gray-box manner, requiring no knowledge about the decoder module. We also show that our attacks fool the popular open-source platform Hugging Face.
Authors: Taimeng Fu, Shaoshu Su, Chen Wang
Simultaneous localization and mapping (SLAM) stands as one of the critical challenges in robot navigation. Recent advancements suggest that methods based on supervised learning deliver impressive performance in front-end odometry, while traditional optimization-based methods still play a vital role in the back-end for minimizing estimation drift. In this paper, we found that such decoupled paradigm can lead to only sub-optimal performance, consequently curtailing system capabilities and generalization potential. To solve this problem, we proposed a novel self-supervised learning framework, imperative SLAM (iSLAM), which fosters reciprocal correction between the front-end and back-end, thus enhancing performance without necessitating any external supervision. Specifically, we formulate a SLAM system as a bi-level optimization problem so that the two components are bidirectionally connected. As a result, the front-end model is able to learn global geometric knowledge obtained through pose graph optimization by back-propagating the residuals from the back-end. This significantly improves the generalization ability of the entire system and thus achieves the accuracy improvement up to 45%. To the best of our knowledge, iSLAM is the first SLAM system showing that the front-end and back-end can learn jointly and mutually contribute to each other in a self-supervised manner.
Authors: Mahyar Khayatkhoei, Wael AbdAlmageed
Precision and Recall are two prominent metrics of generative performance, which were proposed to separately measure the fidelity and diversity of generative models. Given their central role in comparing and improving generative models, understanding their limitations are crucially important. To that end, in this work, we identify a critical flaw in the common approximation of these metrics using k-nearest-neighbors, namely, that the very interpretations of fidelity and diversity that are assigned to Precision and Recall can fail in high dimensions, resulting in very misleading conclusions. Specifically, we empirically and theoretically show that as the number of dimensions grows, two model distributions with supports at equal point-wise distance from the support of the real distribution, can have vastly different Precision and Recall regardless of their respective distributions, hence an emergent asymmetry in high dimensions. Based on our theoretical insights, we then provide simple yet effective modifications to these metrics to construct symmetric metrics regardless of the number of dimensions. Finally, we provide experiments on real-world datasets to illustrate that the identified flaw is not merely a pathological case, and that our proposed metrics are effective in alleviating its impact.
Authors: Hsiang-Wei Huang, Cheng-Yen Yang, Jiacheng Sun, Jenq-Neng Hwang, Chung-I Huang
Multi-object tracking algorithms have made significant advancements due to the recent developments in object detection. However, most existing methods primarily focus on tracking pedestrians or vehicles, which exhibit relatively simple and regular motion patterns. Consequently, there is a scarcity of algorithms that address the tracking of targets with irregular or non-linear motion, such as multi-athlete tracking. Furthermore, popular tracking algorithms often rely on the Kalman filter for object motion modeling, which fails to track objects when their motion contradicts the linear motion assumption of the Kalman filter. Due to this reason, we proposed a novel online and robust multi-object tracking approach, named Iterative Scale-Up ExpansionIoU and Deep Features for multi-object tracking. Unlike conventional methods, we abandon the use of the Kalman filter and propose utilizing the iterative scale-up expansion IoU. This approach achieves superior tracking performance without requiring additional training data or adopting a more robust detector, all while maintaining a lower computational cost compared to other appearance-based methods. Our proposed method demonstrates remarkable effectiveness in tracking irregular motion objects, achieving a score of 76.9% in HOTA. It outperforms all state-of-the-art tracking algorithms on the SportsMOT dataset, covering various kinds of sport scenarios.
Authors: Mingyuan Luo, Xin Yang, Zhongnuo Yan, Junyu Li, Yuanji Zhang, Jiongquan Chen, Xindi Hu, Jikuan Qian, Jun Cheng, Dong Ni
Ultrasound (US) imaging is a popular tool in clinical diagnosis, offering safety, repeatability, and real-time capabilities. Freehand 3D US is a technique that provides a deeper understanding of scanned regions without increasing complexity. However, estimating elevation displacement and accumulation error remains challenging, making it difficult to infer the relative position using images alone. The addition of external lightweight sensors has been proposed to enhance reconstruction performance without adding complexity, which has been shown to be beneficial. We propose a novel online self-consistency network (OSCNet) using multiple inertial measurement units (IMUs) to improve reconstruction performance. OSCNet utilizes a modal-level self-supervised strategy to fuse multiple IMU information and reduce differences between reconstruction results obtained from each IMU data. Additionally, a sequence-level self-consistency strategy is proposed to improve the hierarchical consistency of prediction results among the scanning sequence and its sub-sequences. Experiments on large-scale arm and carotid datasets with multiple scanning tactics demonstrate that our OSCNet outperforms previous methods, achieving state-of-the-art reconstruction performance.
Authors: Tserendorj Adiya, Sanghun Kim, Jung Eun Lee, Jae Shin Yoon, Hwasup Lim
We introduce a method to generate temporally coherent human animation from a single image, a video, or a random noise. This problem has been formulated as modeling of an auto-regressive generation, i.e., to regress past frames to decode future frames. However, such unidirectional generation is highly prone to motion drifting over time, generating unrealistic human animation with significant artifacts such as appearance distortion. We claim that bidirectional temporal modeling enforces temporal coherence on a generative network by largely suppressing the motion ambiguity of human appearance. To prove our claim, we design a novel human animation framework using a denoising diffusion model: a neural network learns to generate the image of a person by denoising temporal Gaussian noises whose intermediate results are cross-conditioned bidirectionally between consecutive frames. In the experiments, our method demonstrates strong performance compared to existing unidirectional approaches with realistic temporal coherence
Authors: Anil Osman Tur, Nicola Dall'Asen, Cigdem Beyan, Elisa Ricci
This paper aims to address the unsupervised video anomaly detection (VAD) problem, which involves classifying each frame in a video as normal or abnormal, without any access to labels. To accomplish this, the proposed method employs conditional diffusion models, where the input data is the spatiotemporal features extracted from a pre-trained network, and the condition is the features extracted from compact motion representations that summarize a given video segment in terms of its motion and appearance. Our method utilizes a data-driven threshold and considers a high reconstruction error as an indicator of anomalous events. This study is the first to utilize compact motion representations for VAD and the experiments conducted on two large-scale VAD benchmarks demonstrate that they supply relevant information to the diffusion model, and consequently improve VAD performances w.r.t the prior art. Importantly, our method exhibits better generalization performance across different datasets, notably outperforming both the state-of-the-art and baseline methods. The code of our method is available at https://github.com/AnilOsmanTur/conditioned_video_anomaly_diffusion
Authors: Fatemeh Farokhmanesh, Kevin Höhlein, Christoph Neuhauser, Rüdiger Westermann
We present the first neural network that has learned to compactly represent and can efficiently reconstruct the statistical dependencies between the values of physical variables at different spatial locations in large 3D simulation ensembles. Going beyond linear dependencies, we consider mutual information as a measure of non-linear dependence. We demonstrate learning and reconstruction with a large weather forecast ensemble comprising 1000 members, each storing multiple physical variables at a 250 x 352 x 20 simulation grid. By circumventing compute-intensive statistical estimators at runtime, we demonstrate significantly reduced memory and computation requirements for reconstructing the major dependence structures. This enables embedding the estimator into a GPU-accelerated direct volume renderer and interactively visualizing all mutual dependencies for a selected domain point.
Authors: Xuanlin Li, Yunhao Fang, Minghua Liu, Zhan Ling, Zhuowen Tu, Hao Su
Large vision-language models have achieved outstanding performance, but their size and computational requirements make their deployment on resource-constrained devices and time-sensitive tasks impractical. Model distillation, the process of creating smaller, faster models that maintain the performance of larger models, is a promising direction towards the solution. This paper investigates the distillation of visual representations in large teacher vision-language models into lightweight student models using a small- or mid-scale dataset. Notably, this study focuses on open-vocabulary out-of-distribution (OOD) generalization, a challenging problem that has been overlooked in previous model distillation literature. We propose two principles from vision and language modality perspectives to enhance student's OOD generalization: (1) by better imitating teacher's visual representation space, and carefully promoting better coherence in vision-language alignment with the teacher; (2) by enriching the teacher's language representations with informative and finegrained semantic attributes to effectively distinguish between different labels. We propose several metrics and conduct extensive experiments to investigate their techniques. The results demonstrate significant improvements in zero-shot and few-shot student performance on open-vocabulary out-of-distribution classification, highlighting the effectiveness of our proposed approaches. Code released at https://github.com/xuanlinli17/large_vlm_distillation_ood
Authors: Kyriaki-Margarita Bintsi, Vasileios Baltatzis, Rolandos Alexandros Potamias, Alexander Hammers, Daniel Rueckert
Brain age estimation is clinically important as it can provide valuable information in the context of neurodegenerative diseases such as Alzheimer's. Population graphs, which include multimodal imaging information of the subjects along with the relationships among the population, have been used in literature along with Graph Convolutional Networks (GCNs) and have proved beneficial for a variety of medical imaging tasks. A population graph is usually static and constructed manually using non-imaging information. However, graph construction is not a trivial task and might significantly affect the performance of the GCN, which is inherently very sensitive to the graph structure. In this work, we propose a framework that learns a population graph structure optimized for the downstream task. An attention mechanism assigns weights to a set of imaging and non-imaging features (phenotypes), which are then used for edge extraction. The resulting graph is used to train the GCN. The entire pipeline can be trained end-to-end. Additionally, by visualizing the attention weights that were the most important for the graph construction, we increase the interpretability of the graph. We use the UK Biobank, which provides a large variety of neuroimaging and non-imaging phenotypes, to evaluate our method on brain age regression and classification. The proposed method outperforms competing static graph approaches and other state-of-the-art adaptive methods. We further show that the assigned attention scores indicate that there are both imaging and non-imaging phenotypes that are informative for brain age estimation and are in agreement with the relevant literature.
Authors: Rakshith Subramanyam, T. S. Jayram, Rushil Anirudh, Jayaraman J. Thiagarajan
In this paper, we explore the potential of Vision-Language Models (VLMs), specifically CLIP, in predicting visual object relationships, which involves interpreting visual features from images into language-based relations. Current state-of-the-art methods use complex graphical models that utilize language cues and visual features to address this challenge. We hypothesize that the strong language priors in CLIP embeddings can simplify these graphical models paving for a simpler approach. We adopt the UVTransE relation prediction framework, which learns the relation as a translational embedding with subject, object, and union box embeddings from a scene. We systematically explore the design of CLIP-based subject, object, and union-box representations within the UVTransE framework and propose CREPE (CLIP Representation Enhanced Predicate Estimation). CREPE utilizes text-based representations for all three bounding boxes and introduces a novel contrastive training strategy to automatically infer the text prompt for union-box. Our approach achieves state-of-the-art performance in predicate estimation, mR@5 27.79, and mR@20 31.95 on the Visual Genome benchmark, achieving a 15.3\% gain in performance over recent state-of-the-art at mR@20. This work demonstrates CLIP's effectiveness in object relation prediction and encourages further research on VLMs in this challenging domain.
Authors: Fadi Boutros, Marcel Klemt, Meiling Fang, Arjan Kuijper, Naser Damer
Deep generative models have recently presented impressive results in generating realistic face images of random synthetic identities.
To generate multiple samples of a certain synthetic identity, previous works proposed to disentangle the latent space of GANs by incorporating additional supervision or regularization, enabling the manipulation of certain attributes. Others proposed to disentangle specific factors in unconditional pretrained GANs latent spaces to control their output, which also requires supervision by attribute classifiers. Moreover, these attributes are entangled in GAN's latent space, making it difficult to manipulate them without affecting the identity information. We propose in this work a framework, ExFaceGAN, to disentangle identity information in pretrained GANs latent spaces, enabling the generation of multiple samples of any synthetic identity. Given a reference latent code of any synthetic image and latent space of pretrained GAN, our ExFaceGAN learns an identity directional boundary that disentangles the latent space into two sub-spaces, with latent codes of samples that are either identity similar or dissimilar to a reference image. By sampling from each side of the boundary, our ExFaceGAN can generate multiple samples of synthetic identity without the need for designing a dedicated architecture or supervision from attribute classifiers. We demonstrate the generalizability and effectiveness of ExFaceGAN by integrating it into learned latent spaces of three SOTA GAN approaches. As an example of the practical benefit of our ExFaceGAN, we empirically prove that data generated by ExFaceGAN can be successfully used to train face recognition models (\url{https://github.com/fdbtrs/ExFaceGAN}).
Authors: Kalyan Ramakrishnan
Audio-Visual Event Localization (AVEL) is the task of temporally localizing and classifying \emph{audio-visual events}, i.e., events simultaneously visible and audible in a video. In this paper, we solve AVEL in a weakly-supervised setting, where only video-level event labels (their presence/absence, but not their locations in time) are available as supervision for training. Our idea is to use a base model to estimate labels on the training data at a finer temporal resolution than at the video level and re-train the model with these labels. I.e., we determine the subset of labels for each \emph{slice} of frames in a training video by (i) replacing the frames outside the slice with those from a second video having no overlap in video-level labels, and (ii) feeding this synthetic video into the base model to extract labels for just the slice in question. To handle the out-of-distribution nature of our synthetic videos, we propose an auxiliary objective for the base model that induces more reliable predictions of the localized event labels as desired. Our three-stage pipeline outperforms several existing AVEL methods with no architectural changes and improves performance on a related weakly-supervised task as well.
Authors: Kai Su, Qiangfu Zhao, Yoichi Tomioka, Yong Liu
In the realm of Tiny AI, we introduce "You Only Look at Interested Cells" (YOLIC), an efficient method for object localization and classification on edge devices. Seamlessly blending the strengths of semantic segmentation and object detection, YOLIC offers superior computational efficiency and precision. By adopting Cells of Interest for classification instead of individual pixels, YOLIC encapsulates relevant information, reduces computational load, and enables rough object shape inference. Importantly, the need for bounding box regression is obviated, as YOLIC capitalizes on the predetermined cell configuration that provides information about potential object location, size, and shape. To tackle the issue of single-label classification limitations, a multi-label classification approach is applied to each cell, effectively recognizing overlapping or closely situated objects. This paper presents extensive experiments on multiple datasets, demonstrating that YOLIC achieves detection performance comparable to the state-of-the-art YOLO algorithms while surpassing in speed, exceeding 30fps on a Raspberry Pi 4B CPU. All resources related to this study, including datasets, cell designer, image annotation tool, and source code, have been made publicly available on our project website at https://kai3316.github.io/yolic.github.io
Authors: Pietro Bonazzi, Thomas Ruegg, Sizhen Bian, Yawei Li, Michele Magno
Intelligent edge vision tasks encounter the critical challenge of ensuring power and latency efficiency due to the typically heavy computational load they impose on edge platforms.This work leverages one of the first "AI in sensor" vision platforms, IMX500 by Sony, to achieve ultra-fast and ultra-low-power end-to-end edge vision applications. We evaluate the IMX500 and compare it to other edge platforms, such as the Google Coral Dev Micro and Sony Spresense, by exploring gaze estimation as a case study. We propose TinyTracker, a highly efficient, fully quantized model for 2D gaze estimation designed to maximize the performance of the edge vision systems considered in this study. TinyTracker achieves a 41x size reduction (600Kb) compared to iTracker [1] without significant loss in gaze estimation accuracy (maximum of 0.16 cm when fully quantized). TinyTracker's deployment on the Sony IMX500 vision sensor results in end-to-end latency of around 19ms. The camera takes around 17.9ms to read, process and transmit the pixels to the accelerator. The inference time of the network is 0.86ms with an additional 0.24 ms for retrieving the results from the sensor. The overall energy consumption of the end-to-end system is 4.9 mJ, including 0.06 mJ for inference. The end-to-end study shows that IMX500 is 1.7x faster than CoralMicro (19ms vs 34.4ms) and 7x more power efficient (4.9mJ VS 34.2mJ)
Authors: Xingxing Wei, Yao Huang, Yitong Sun, Jie Yu
Recently, physical adversarial attacks have been presented to evade DNNs-based object detectors. To ensure the security, many scenarios are simultaneously deployed with visible sensors and infrared sensors, leading to the failures of these single-modal physical attacks. To show the potential risks under such scenes, we propose a unified adversarial patch to perform cross-modal physical attacks, i.e., fooling visible and infrared object detectors at the same time via a single patch. Considering different imaging mechanisms of visible and infrared sensors, our work focuses on modeling the shapes of adversarial patches, which can be captured in different modalities when they change. To this end, we design a novel boundary-limited shape optimization to achieve the compact and smooth shapes, and thus they can be easily implemented in the physical world. In addition, to balance the fooling degree between visible detector and infrared detector during the optimization process, we propose a score-aware iterative evaluation, which can guide the adversarial patch to iteratively reduce the predicted scores of the multi-modal sensors. We finally test our method against the one-stage detector: YOLOv3 and the two-stage detector: Faster RCNN. Results show that our unified patch achieves an Attack Success Rate (ASR) of 73.33% and 69.17%, respectively. More importantly, we verify the effective attacks in the physical world when visible and infrared sensors shoot the objects under various settings like different angles, distances, postures, and scenes.
Authors: Yechao Zhang, Shengshan Hu, Leo Yu Zhang, Junyu Shi, Minghui Li, Xiaogeng Liu, Wei Wan, Hai Jin
Adversarial examples (AEs) for DNNs have been shown to be transferable: AEs that successfully fool white-box surrogate models can also deceive other black-box models with different architectures. Although a bunch of empirical studies have provided guidance on generating highly transferable AEs, many of these findings lack explanations and even lead to inconsistent advice. In this paper, we take a further step towards understanding adversarial transferability, with a particular focus on surrogate aspects. Starting from the intriguing little robustness phenomenon, where models adversarially trained with mildly perturbed adversarial samples can serve as better surrogates, we attribute it to a trade-off between two predominant factors: model smoothness and gradient similarity. Our investigations focus on their joint effects, rather than their separate correlations with transferability. Through a series of theoretical and empirical analyses, we conjecture that the data distribution shift in adversarial training explains the degradation of gradient similarity. Building on these insights, we explore the impacts of data augmentation and gradient regularization on transferability and identify that the trade-off generally exists in the various training mechanisms, thus building a comprehensive blueprint for the regulation mechanism behind transferability. Finally, we provide a general route for constructing better surrogates to boost transferability which optimizes both model smoothness and gradient similarity simultaneously, e.g., the combination of input gradient regularization and sharpness-aware minimization (SAM), validated by extensive experiments. In summary, we call for attention to the united impacts of these two factors for launching effective transfer attacks, rather than optimizing one while ignoring the other, and emphasize the crucial role of manipulating surrogate models.
Authors: Xiaohang Ren, Xingyu Chen, Pengfei Yao, Heung-Yeung Shum, Baoyuan Wang
The SOTA face swap models still suffer the problem of either target identity (i.e., shape) being leaked or the target non-identity attributes (i.e., background, hair) failing to be fully preserved in the final results. We show that this insufficient disentanglement is caused by two flawed designs that were commonly adopted in prior models: (1) counting on only one compressed encoder to represent both the semantic-level non-identity facial attributes(i.e., pose) and the pixel-level non-facial region details, which is contradictory to satisfy at the same time; (2) highly relying on long skip-connections between the encoder and the final generator, leaking a certain amount of target face identity into the result. To fix them, we introduce a new face swap framework called 'WSC-swap' that gets rid of skip connections and uses two target encoders to respectively capture the pixel-level non-facial region attributes and the semantic non-identity attributes in the face region. To further reinforce the disentanglement learning for the target encoder, we employ both identity removal loss via adversarial training (i.e., GAN) and the non-identity preservation loss via prior 3DMM models like [11]. Extensive experiments on both FaceForensics++ and CelebA-HQ show that our results significantly outperform previous works on a rich set of metrics, including one novel metric for measuring identity consistency that was completely neglected before.
Authors: Jinlong Li, Runsheng Xu, Xinyu Liu, Baolu Li, Qin Zou, Jiaqi Ma, Hongkai Yu
Due to the lack of real multi-agent data and time-consuming of labeling, existing multi-agent cooperative perception algorithms usually select the simulated sensor data for training and validating. However, the perception performance is degraded when these simulation-trained models are deployed to the real world, due to the significant domain gap between the simulated and real data. In this paper, we propose the first Simulation-to-Reality transfer learning framework for multi-agent cooperative perception using a novel Vision Transformer, named as S2R-ViT, which considers both the Implementation Gap and Feature Gap between simulated and real data. We investigate the effects of these two types of domain gaps and propose a novel uncertainty-aware vision transformer to effectively relief the Implementation Gap and an agent-based feature adaptation module with inter-agent and ego-agent discriminators to reduce the Feature Gap. Our intensive experiments on the public multi-agent cooperative perception datasets OPV2V and V2V4Real demonstrate that the proposed S2R-ViT can effectively bridge the gap from simulation to reality and outperform other methods significantly for point cloud-based 3D object detection.
Authors: Yujiao Shi, Fei Wu, Ankit Vora, Akhil Perincherry, Hongdong Li
Image retrieval-based cross-view localization methods often lead to very coarse camera pose estimation, due to the limited sampling density of the database satellite images. In this paper, we propose a method to increase the accuracy of a ground camera's location and orientation by estimating the relative rotation and translation between the ground-level image and its matched/retrieved satellite image. Our approach designs a geometry-guided cross-view transformer that combines the benefits of conventional geometry and learnable cross-view transformers to map the ground-view observations to an overhead view. Given the synthesized overhead view and observed satellite feature maps, we construct a neural pose optimizer with strong global information embedding ability to estimate the relative rotation between them. After aligning their rotations, we develop an uncertainty-guided spatial correlation to generate a probability map of the vehicle locations, from which the relative translation can be determined. Experimental results demonstrate that our method significantly outperforms the state-of-the-art. Notably, the likelihood of restricting the vehicle lateral pose to be within 1m of its Ground Truth (GT) value on the cross-view KITTI dataset has been improved from $35.54\%$ to $76.44\%$, and the likelihood of restricting the vehicle orientation to be within $1^{\circ}$ of its GT value has been improved from $19.64\%$ to $99.10\%$.
Authors: Che Liu, Sibo Cheng, Chen Chen, Mengyun Qiao, Weitong Zhang, Anand Shah, Wenjia Bai, Rossella Arcucci
Medical vision-language models enable co-learning and integrating features from medical imaging and clinical text. However, these models are not easy to train and the latent representation space can be complex. Here we propose a novel way for pre-training and regularising medical vision-language models. The proposed method, named Medical vision-language pre-training with Frozen language models and Latent spAce Geometry optimization (M-FLAG), leverages a frozen language model for training stability and efficiency and introduces a novel orthogonality loss to harmonize the latent space geometry. We demonstrate the potential of the pre-trained model on three downstream tasks: medical image classification, segmentation, and object detection. Extensive experiments across five public datasets demonstrate that M-FLAG significantly outperforms existing medical vision-language pre-training approaches and reduces the number of parameters by 78\%. Notably, M-FLAG achieves outstanding performance on the segmentation task while using only 1\% of the RSNA dataset, even outperforming ImageNet pre-trained models that have been fine-tuned using 100\% of the data.
Authors: Qing Jiang, Jiapeng Wang, Dezhi Peng, Chongyu Liu, Lianwen Jin
This paper aims to re-assess scene text recognition (STR) from a data-oriented perspective. We begin by revisiting the six commonly used benchmarks in STR and observe a trend of performance saturation, whereby only 2.91% of the benchmark images cannot be accurately recognized by an ensemble of 13 representative models. While these results are impressive and suggest that STR could be considered solved, however, we argue that this is primarily due to the less challenging nature of the common benchmarks, thus concealing the underlying issues that STR faces. To this end, we consolidate a large-scale real STR dataset, namely Union14M, which comprises 4 million labeled images and 10 million unlabeled images, to assess the performance of STR models in more complex real-world scenarios. Our experiments demonstrate that the 13 models can only achieve an average accuracy of 66.53% on the 4 million labeled images, indicating that STR still faces numerous challenges in the real world. By analyzing the error patterns of the 13 models, we identify seven open challenges in STR and develop a challenge-driven benchmark consisting of eight distinct subsets to facilitate further progress in the field. Our exploration demonstrates that STR is far from being solved and leveraging data may be a promising solution. In this regard, we find that utilizing the 10 million unlabeled images through self-supervised pre-training can significantly improve the robustness of STR model in real-world scenarios and leads to state-of-the-art performance.
Authors: Rundong Luo, Wenjing Wang, Wenhan Yang, Jiaying Liu
Low-light conditions not only hamper human visual experience but also degrade the model's performance on downstream vision tasks. While existing works make remarkable progress on day-night domain adaptation, they rely heavily on domain knowledge derived from the task-specific nighttime dataset. This paper challenges a more complicated scenario with border applicability, i.e., zero-shot day-night domain adaptation, which eliminates reliance on any nighttime data. Unlike prior zero-shot adaptation approaches emphasizing either image-level translation or model-level adaptation, we propose a similarity min-max paradigm that considers them under a unified framework. On the image level, we darken images towards minimum feature similarity to enlarge the domain gap. Then on the model level, we maximize the feature similarity between the darkened images and their normal-light counterparts for better model adaptation. To the best of our knowledge, this work represents the pioneering effort in jointly optimizing both aspects, resulting in a significant improvement of model generalizability. Extensive experiments demonstrate our method's effectiveness and broad applicability on various nighttime vision tasks, including classification, semantic segmentation, visual place recognition, and video action recognition. Code and pre-trained models are available at https://red-fairy.github.io/ZeroShotDayNightDA-Webpage/.
Authors: Zeen Song, Xingzhe Su, Jingyao Wang, Wenwen Qiang, Changwen Zheng, Fuchun Sun
In recent years, self-supervised learning (SSL) has emerged as a promising approach for extracting valuable representations from unlabeled data. One successful SSL method is contrastive learning, which aims to bring positive examples closer while pushing negative examples apart. Many current contrastive learning approaches utilize a parameterized projection head. Through a combination of empirical analysis and theoretical investigation, we provide insights into the internal mechanisms of the projection head and its relationship with the phenomenon of dimensional collapse. Our findings demonstrate that the projection head enhances the quality of representations by performing contrastive loss in a projected subspace. Therefore, we propose an assumption that only a subset of features is necessary when minimizing the contrastive loss of a mini-batch of data. Theoretical analysis further suggests that a sparse projection head can enhance generalization, leading us to introduce SparseHead - a regularization term that effectively constrains the sparsity of the projection head, and can be seamlessly integrated with any self-supervised learning (SSL) approaches. Our experimental results validate the effectiveness of SparseHead, demonstrating its ability to improve the performance of existing contrastive methods.
Authors: Zhiyu Wu, Jinshi Cui
Facial expression recognition (FER) remains a challenging task due to the ambiguity of expressions. The derived noisy labels significantly harm the performance in real-world scenarios. To address this issue, we present a new FER model named Landmark-Aware Net~(LA-Net), which leverages facial landmarks to mitigate the impact of label noise from two perspectives. Firstly, LA-Net uses landmark information to suppress the uncertainty in expression space and constructs the label distribution of each sample by neighborhood aggregation, which in turn improves the quality of training supervision. Secondly, the model incorporates landmark information into expression representations using the devised expression-landmark contrastive loss. The enhanced expression feature extractor can be less susceptible to label noise. Our method can be integrated with any deep neural network for better training supervision without introducing extra inference costs. We conduct extensive experiments on both in-the-wild datasets and synthetic noisy datasets and demonstrate that LA-Net achieves state-of-the-art performance.
Authors: Shaoxu Li
We propose a method for synthesizing photo-realistic digital avatars from only one portrait as the reference. Given a portrait, our method synthesizes a coarse talking head video using driving keypoints features. And with the coarse video, our method synthesizes a coarse talking head avatar with a deforming neural radiance field. With rendered images of the coarse avatar, our method updates the low-quality images with a blind face restoration model. With updated images, we retrain the avatar for higher quality. After several iterations, our method can synthesize a photo-realistic animatable 3D neural head avatar. The motivation of our method is deformable neural radiance field can eliminate the unnatural distortion caused by the image2video method. Our method outperforms state-of-the-art methods in quantitative and qualitative studies on various subjects.
Authors: Zhixiang Wei, Lin Chen, Tao Tu, Huaian Chen, Pengyang Ling, Yi Jin
Most prior semantic segmentation methods have been developed for day-time scenes, while typically underperforming in night-time scenes due to insufficient and complicated lighting conditions. In this work, we tackle this challenge by proposing a novel night-time semantic segmentation paradigm, i.e., disentangle then parse (DTP). DTP explicitly disentangles night-time images into light-invariant reflectance and light-specific illumination components and then recognizes semantics based on their adaptive fusion. Concretely, the proposed DTP comprises two key components: 1) Instead of processing lighting-entangled features as in prior works, our Semantic-Oriented Disentanglement (SOD) framework enables the extraction of reflectance component without being impeded by lighting, allowing the network to consistently recognize the semantics under cover of varying and complicated lighting conditions. 2) Based on the observation that the illumination component can serve as a cue for some semantically confused regions, we further introduce an Illumination-Aware Parser (IAParser) to explicitly learn the correlation between semantics and lighting, and aggregate the illumination features to yield more precise predictions. Extensive experiments on the night-time segmentation task with various settings demonstrate that DTP significantly outperforms state-of-the-art methods. Furthermore, with negligible additional parameters, DTP can be directly used to benefit existing day-time methods for night-time segmentation.
Authors: Federico Betti, Jacopo Staiano, Lorenzo Baraldi, Lorenzo Baraldi, Rita Cucchiara, Nicu Sebe
Research in Image Generation has recently made significant progress, particularly boosted by the introduction of Vision-Language models which are able to produce high-quality visual content based on textual inputs. Despite ongoing advancements in terms of generation quality and realism, no methodical frameworks have been defined yet to quantitatively measure the quality of the generated content and the adherence with the prompted requests: so far, only human-based evaluations have been adopted for quality satisfaction and for comparing different generative methods. We introduce a novel automated method for Visual Concept Evaluation (ViCE), i.e. to assess consistency between a generated/edited image and the corresponding prompt/instructions, with a process inspired by the human cognitive behaviour. ViCE combines the strengths of Large Language Models (LLMs) and Visual Question Answering (VQA) into a unified pipeline, aiming to replicate the human cognitive process in quality assessment. This method outlines visual concepts, formulates image-specific verification questions, utilizes the Q&A system to investigate the image, and scores the combined outcome. Although this brave new hypothesis of mimicking humans in the image evaluation process is in its preliminary assessment stage, results are promising and open the door to a new form of automatic evaluation which could have significant impact as the image generation or the image target editing tasks become more and more sophisticated.
Authors: Fatemeh Rezapoor Nikroo, Ajinkya Deshmukh, Anantha Sharma, Adrian Tam, Kaarthik Kumar, Cleo Norris, Aditya Dangi
In this study, we evaluate the performance of multiple state-of-the-art SRGAN (Super Resolution Generative Adversarial Network) models, ESRGAN, Real-ESRGAN and EDSR, on a benchmark dataset of real-world images which undergo degradation using a pipeline. Our results show that some models seem to significantly increase the resolution of the input images while preserving their visual quality, this is assessed using Tesseract OCR engine. We observe that EDSR-BASE model from huggingface outperforms the remaining candidate models in terms of both quantitative metrics and subjective visual quality assessments with least compute overhead. Specifically, EDSR generates images with higher peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM) values and are seen to return high quality OCR results with Tesseract OCR engine. These findings suggest that EDSR is a robust and effective approach for single-image super-resolution and may be particularly well-suited for applications where high-quality visual fidelity is critical and optimized compute.