Authors: Yifan Shen, Yuanzhe Liu, Jingyuan Zhu, Xu Cao, Xiaofeng Zhang, Yixiao He, Wenming Ye, James Matthew Rehg, Ismini Lourentzou
Abstract: Current Vision-Language Models (VLMs) struggle with fine-grained spatial reasoning, particularly when multi-step logic and precise spatial alignment are required. In this work, we introduce SpatialReasoner-R1, a vision-language reasoning model designed to address these limitations. To construct high-quality supervision for spatial reasoning, we design a Multi-Model Monte Carlo Tree Search (M3CTS) method that generates diverse, logically consistent Long Chain-of-Thought (LongCoT) reasoning trajectories. In addition, we propose fine-grained Direct Preference Optimization (fDPO), which introduces segment-specific preference granularity for descriptive grounding and logical reasoning, guided by a spatial reward mechanism that evaluates candidate responses based on visual consistency, spatial grounding, and logical coherence. Experimental results demonstrate that fDPO achieves an average improvement of 4.1% over standard DPO across spatial quality tasks, and a 9.0% gain in spatial quantity tasks. SpatialReasoner-R1, trained with fDPO, sets a new SoTA on SPATIALRGPT-Bench, outperforming the strongest baseline by 9.8% in average accuracy, while maintaining competitive performance on general vision-language tasks.
Authors: Hakan \c{C}apuk, Andrew Bond, Muhammed Burak K{\i}z{\i}l, Emir G\"o\c{c}en, Erkut Erdem, Aykut Erdem
Abstract: Recent advances in image generation have led to remarkable improvements in synthesizing perspective images. However, these models still struggle with panoramic image generation due to unique challenges, including varying levels of geometric distortion and the requirement for seamless loop-consistency. To address these issues while leveraging the strengths of the existing models, we introduce TanDiT, a method that synthesizes panoramic scenes by generating grids of tangent-plane images covering the entire 360$^\circ$ view. Unlike previous methods relying on multiple diffusion branches, TanDiT utilizes a unified diffusion model trained to produce these tangent-plane images simultaneously within a single denoising iteration. Furthermore, we propose a model-agnostic post-processing step specifically designed to enhance global coherence across the generated panoramas. To accurately assess panoramic image quality, we also present two specialized metrics, TangentIS and TangentFID, and provide a comprehensive benchmark comprising captioned panoramic datasets and standardized evaluation scripts. Extensive experiments demonstrate that our method generalizes effectively beyond its training data, robustly interprets detailed and complex text prompts, and seamlessly integrates with various generative models to yield high-quality, diverse panoramic images.
Authors: Liangyu Zhong, Fabio Rosenthal, Joachim Sicking, Fabian H\"uger, Thorsten Bagdonat, Hanno Gottschalk, Leo Schwinn
Abstract: While Multimodal Large Language Models (MLLMs) offer strong perception and reasoning capabilities for image-text input, Visual Question Answering (VQA) focusing on small image details still remains a challenge. Although visual cropping techniques seem promising, recent approaches have several limitations: the need for task-specific fine-tuning, low efficiency due to uninformed exhaustive search, or incompatibility with efficient attention implementations. We address these shortcomings by proposing a training-free visual cropping method, dubbed FOCUS, that leverages MLLM-internal representations to guide the search for the most relevant image region. This is accomplished in four steps: first, we identify the target object(s) in the VQA prompt; second, we compute an object relevance map using the key-value (KV) cache; third, we propose and rank relevant image regions based on the map; and finally, we perform the fine-grained VQA task using the top-ranked region. As a result of this informed search strategy, FOCUS achieves strong performance across four fine-grained VQA datasets and two types of MLLMs. It outperforms three popular visual cropping methods in both accuracy and efficiency, and matches the best-performing baseline, ZoomEye, while requiring 3 - 6.5 x less compute.
Authors: Aryan Thakre, Omkar Nagwekar, Vedang Talekar, Aparna Santra Biswas
Abstract: Deepfakes have emerged as a significant threat to digital media authenticity, increasing the need for advanced detection techniques that can identify subtle and time-dependent manipulations. CNNs are effective at capturing spatial artifacts, and Transformers excel at modeling temporal inconsistencies. However, many existing CNN-Transformer models process spatial and temporal features independently. In particular, attention-based methods often use separate attention mechanisms for spatial and temporal features and combine them using naive approaches like averaging, addition, or concatenation, which limits the depth of spatio-temporal interaction. To address this challenge, we propose a unified CAST model that leverages cross-attention to effectively fuse spatial and temporal features in a more integrated manner. Our approach allows temporal features to dynamically attend to relevant spatial regions, enhancing the model's ability to detect fine-grained, time-evolving artifacts such as flickering eyes or warped lips. This design enables more precise localization and deeper contextual understanding, leading to improved performance across diverse and challenging scenarios. We evaluate the performance of our model using the FaceForensics++, Celeb-DF, and DeepfakeDetection datasets in both intra- and cross-dataset settings to affirm the superiority of our approach. Our model achieves strong performance with an AUC of 99.49 percent and an accuracy of 97.57 percent in intra-dataset evaluations. In cross-dataset testing, it demonstrates impressive generalization by achieving a 93.31 percent AUC on the unseen DeepfakeDetection dataset. These results highlight the effectiveness of cross-attention-based feature fusion in enhancing the robustness of deepfake video detection.
Authors: Xin Lu, Xueyang Fu, Jie Xiao, Zihao Fan, Yurui Zhu, Zheng-Jun Zha
Abstract: While diffusion models demonstrate strong generative capabilities in image restoration (IR) tasks, their complex architectures and iterative processes limit their practical application compared to mainstream reconstruction-based general ordinary IR networks. Existing approaches primarily focus on optimizing network architecture and diffusion paths but overlook the integration of the diffusion training paradigm within general ordinary IR frameworks. To address these challenges, this paper elucidates key principles for adapting the diffusion training paradigm to general IR training through systematic analysis of time-step dependencies, network hierarchies, noise-level relationships, and multi-restoration task correlations, proposing a new IR framework supported by diffusion-based training. To enable IR networks to simultaneously restore images and model generative representations, we introduce a series of regularization strategies that align diffusion objectives with IR tasks, improving generalization in single-task scenarios. Furthermore, recognizing that diffusion-based generation exerts varying influences across different IR tasks, we develop an incremental training paradigm and task-specific adaptors, further enhancing performance in multi-task unified IR. Experiments demonstrate that our method significantly improves the generalization of IR networks in single-task IR and achieves superior performance in multi-task unified IR. Notably, the proposed framework can be seamlessly integrated into existing general IR architectures.
Authors: Remco F. Leijenaar, Hamidreza Kasaei
Abstract: Learning semantically meaningful representations from unstructured 3D point clouds remains a central challenge in computer vision, especially in the absence of large-scale labeled datasets. While masked point modeling (MPM) is widely used in self-supervised 3D learning, its reconstruction-based objective can limit its ability to capture high-level semantics. We propose AsymDSD, an Asymmetric Dual Self-Distillation framework that unifies masked modeling and invariance learning through prediction in the latent space rather than the input space. AsymDSD builds on a joint embedding architecture and introduces several key design choices: an efficient asymmetric setup, disabling attention between masked queries to prevent shape leakage, multi-mask sampling, and a point cloud adaptation of multi-crop. AsymDSD achieves state-of-the-art results on ScanObjectNN (90.53%) and further improves to 93.72% when pretrained on 930k shapes, surpassing prior methods.
Authors: Chenqiu Zhao, Anup Basu
Abstract: We propose two theoretical frameworks, the Mutually Exclusive Probability Space (MESP) and the Local Correlation Hypothesis (LCH), to explore a potential limitation in probabilistic generative models; namely that learning global distributions leads to memorization rather than generative behavior. MESP emerges from our rethinking of the Variational Autoencoder (VAE). We observe that latent variable distributions in VAE exhibit overlap, which leads to an optimization conflict between the reconstruction loss and KL-divergence loss. A lower bound based on the overlap coefficient is proposed. We refer to this phenomenon as Mutually Exclusive Probability Spaces. Based on MESP, a Binary Latent Autoencoder (BL-AE) is proposed to encode images into binary latent representations. These binary latents are used as the input to our Autoregressive Random Variable Model (ARVM), a modified autoregressive model outputting histograms. Our ARVM achieves competitive FID scores, outperforming state-of-the-art methods on standard datasets. However, such scores reflect memorization rather than generation. To address this issue, we propose the Local Correlation Hypothesis (LCH), which posits that generative capability arising from local correlations among latent variables. Comprehensive experiments and discussions are conducted to validate our frameworks.
Authors: Nick Lemke, Mirko Konstantin, Henry John Krumb, John Kalkhof, Jonathan Stieber, Anirban Mukhopadhyay
Abstract: Federated Learning (FL) is enabling collaborative model training across institutions without sharing sensitive patient data. This approach is particularly valuable in low- and middle-income countries (LMICs), where access to trained medical professionals is limited. However, FL adoption in LMICs faces significant barriers, including limited high-performance computing resources and unreliable internet connectivity. To address these challenges, we introduce FedNCA, a novel FL system tailored for medical image segmentation tasks. FedNCA leverages the lightweight Med-NCA architecture, enabling training on low-cost edge devices, such as widely available smartphones, while minimizing communication costs. Additionally, our encryption-ready FedNCA proves to be suitable for compromised network communication. By overcoming infrastructural and security challenges, FedNCA paves the way for inclusive, efficient, lightweight, and encryption-ready medical imaging solutions, fostering equitable healthcare advancements in resource-constrained regions.
Authors: Sirnam Swetha, Rohit Gupta, Parth Parag Kulkarni, David G Shatwell, Jeffrey A Chan Santiago, Nyle Siddiqui, Joseph Fioresi, Mubarak Shah
Abstract: Video QA has made significant strides by leveraging multimodal learning to align visual and textual modalities. However, current benchmarks overwhelmingly focus on questions answerable through explicit visual content - actions, objects & events directly observable within individual frames or short clips. In contrast, creative and cinematic videos - such as movies, TV shows, and narrative-driven content - employ storytelling techniques that deliberately omit certain depictions, requiring viewers to infer motives, causality, and relationships across discontinuous frames. Humans naturally excel at such implicit reasoning, seamlessly integrating information across time and context to construct coherent narratives. Current VideoQA systems and benchmarks fail to capture this essential dimension of human-like understanding. To bridge this gap, we present ImplicitQA, a novel benchmark specifically designed to test models on implicit reasoning. It comprises 1K meticulously annotated QA pairs derived from 320+ high-quality creative video clips, systematically categorized into key reasoning dimensions: lateral and vertical spatial reasoning, depth and proximity, viewpoint and visibility, motion and trajectory, causal and motivational reasoning, social interactions, physical context, and inferred counting. These annotations are deliberately challenging, crafted by authors ensuring high-quality. Our extensive evaluations on leading VideoQA models reveals performance degradation, underscoring their reliance on surface-level visual cues and highlighting the difficulty of implicit reasoning. Performance variations across models further illustrate the complexity and diversity of the challenges presented by ImplicitQA. By releasing both the dataset and our data collection framework, we aim to stimulate further research and development in the community. https://huggingface.co/datasets/ucf-crcv/ImplicitQA.
Authors: Rishiraj Paul Chowdhury, Nirmit Shekar Karkera
Abstract: Glaucoma is a leading cause of irreversible blindness, but early detection can significantly improve treatment outcomes. Traditional diagnostic methods are often invasive and require specialized equipment. In this work, we present a deep learning pipeline using the EfficientNet-B0 architecture for glaucoma detection from retinal fundus images. Unlike prior studies that rely on single datasets, we sequentially train and fine-tune our model across ACRIMA, ORIGA, and RIM-ONE datasets to enhance generalization. Our experiments show that minimal preprocessing yields higher AUC-ROC compared to more complex enhancements, and our model demonstrates strong discriminative performance on unseen datasets. The proposed pipeline offers a reproducible and scalable approach to early glaucoma detection, supporting its potential clinical utility.
Authors: Daniel Wen
Abstract: In this study, we investigate various computer vision paradigms - supervised learning, unsupervised learning, and prompt fine-tuning - by assessing their ability to understand and interpret egocentric video data. Specifically, we examine Shotluck Holmes (state-of-the-art supervised learning), TAC-SUM (state-of-the-art unsupervised learning), and GPT-4o (a prompt fine-tuned pre-trained model), evaluating their effectiveness in video summarization. Our results demonstrate that current state-of-the-art models perform less effectively on first-person videos compared to third-person videos, highlighting the need for further advancements in the egocentric video domain. Notably, a prompt fine-tuned general-purpose GPT-4o model outperforms these specialized models, emphasizing the limitations of existing approaches in adapting to the unique challenges of first-person perspectives. Although our evaluation is conducted on a small subset of egocentric videos from the Ego-Exo4D dataset due to resource constraints, the primary objective of this research is to provide a comprehensive proof-of-concept analysis aimed at advancing the application of computer vision techniques to first-person videos. By exploring novel methodologies and evaluating their potential, we aim to contribute to the ongoing development of models capable of effectively processing and interpreting egocentric perspectives.
Authors: Felix Holm, G\"ozde \"Unver, Ghazal Ghazaei, Nassir Navab
Abstract: Understanding the intricate workflows of cataract surgery requires modeling complex interactions between surgical tools, anatomical structures, and procedural techniques. Existing datasets primarily address isolated aspects of surgical analysis, such as tool detection or phase segmentation, but lack comprehensive representations that capture the semantic relationships between entities over time. This paper introduces the Cataract Surgery Scene Graph (CAT-SG) dataset, the first to provide structured annotations of tool-tissue interactions, procedural variations, and temporal dependencies. By incorporating detailed semantic relations, CAT-SG offers a holistic view of surgical workflows, enabling more accurate recognition of surgical phases and techniques. Additionally, we present a novel scene graph generation model, CatSGG, which outperforms current methods in generating structured surgical representations. The CAT-SG dataset is designed to enhance AI-driven surgical training, real-time decision support, and workflow analysis, paving the way for more intelligent, context-aware systems in clinical practice.
Authors: Rafael Sterzinger, Marco Peer, Robert Sablatnig
Abstract: As rich sources of history, maps provide crucial insights into historical changes, yet their diverse visual representations and limited annotated data pose significant challenges for automated processing. We propose a simple yet effective approach for few-shot segmentation of historical maps, leveraging the rich semantic embeddings of large vision foundation models combined with parameter-efficient fine-tuning. Our method outperforms the state-of-the-art on the Siegfried benchmark dataset in vineyard and railway segmentation, achieving +5% and +13% relative improvements in mIoU in 10-shot scenarios and around +20% in the more challenging 5-shot setting. Additionally, it demonstrates strong performance on the ICDAR 2021 competition dataset, attaining a mean PQ of 67.3% for building block segmentation, despite not being optimized for this shape-sensitive metric, underscoring its generalizability. Notably, our approach maintains high performance even in extremely low-data regimes (10- & 5-shot), while requiring only 689k trainable parameters - just 0.21% of the total model size. Our approach enables precise segmentation of diverse historical maps while drastically reducing the need for manual annotations, advancing automated processing and analysis in the field. Our implementation is publicly available at: https://github.com/RafaelSterzinger/few-shot-map-segmentation.
URLs: https://github.com/RafaelSterzinger/few-shot-map-segmentation.
Authors: Minh-Loi Nguyen, Quang-Khai Le, Tam V. Nguyen, Minh-Triet Tran, Trung-Nghia Le
Abstract: Storytelling is a deeply personal and creative process, yet existing methods often treat users as passive consumers, offering generic plots with limited personalization. This undermines engagement and immersion, especially where individual style or appearance is crucial. We introduce TaleForge, a personalized story-generation system that integrates large language models (LLMs) and text-to-image diffusion to embed users' facial images within both narratives and illustrations. TaleForge features three interconnected modules: Story Generation, where LLMs create narratives and character descriptions from user prompts; Personalized Image Generation, merging users' faces and outfit choices into character illustrations; and Background Generation, creating scene backdrops that incorporate personalized characters. A user study demonstrated heightened engagement and ownership when individuals appeared as protagonists. Participants praised the system's real-time previews and intuitive controls, though they requested finer narrative editing tools. TaleForge advances multimodal storytelling by aligning personalized text and imagery to create immersive, user-centric experiences.
Authors: Duy-Bao Bui, Hoang-Khang Nguyen, Trung-Nghia Le
Abstract: Inpainting, the process of filling missing or corrupted image parts, has broad applications, including medical imaging. However, in specialized fields like medical polyps imaging, where accuracy and reliability are critical, inpainting models can generate inaccurate images, leading to significant errors in medical diagnosis and treatment. To ensure reliability, medical images should be annotated by experts like oncologists for effective model training. We propose PrefPaint, an approach that incorporates human feedback into the training process of Stable Diffusion Inpainting, bypassing the need for computationally expensive reward models. In addition, we develop a web-based interface streamlines training, fine-tuning, and inference. This interactive interface provides a smooth and intuitive user experience, making it easier to offer feedback and manage the fine-tuning process. User study on various domains shows that PrefPaint outperforms existing methods, reducing visual inconsistencies and improving image rendering, particularly in medical contexts, where our model generates more realistic polyps images.
Authors: Xiaoqi Wang, Clint Sebastian, Wenbin He, Liu Ren
Abstract: The recent advancements in large foundation models have driven the success of open-set image segmentation, a task focused on segmenting objects beyond predefined categories. Among various prompt types (such as points, boxes, texts, and visual references), visual reference segmentation stands out for its unique flexibility and strong zero-shot capabilities. Recently, several SAM-based methods have made notable progress in this task by automatically generating prompts to guide SAM. However, these methods often generate prompts at object boundaries due to suboptimal prompt encoder, which results in instability and reduced robustness. In this work, we introduce ProSAM, a simple but effective method to address the stability challenges we identified in existing SAM-based visual reference segmentation approaches. By learning a variational prompt encoder to predict multivariate prompt distributions, ProSAM avoids generating prompts that lie in unstable regions, overcoming the instability caused by less robust prompts. Our approach consistently surpasses state-of-the-art methods on the Pascal-5$^i$ and COCO-20$^i$ datasets, providing a more robust solution for visual reference segmentation.
Authors: Mengyi Shan, Brian Curless, Ira Kemelmacher-Shlizerman, Steve Seitz
Abstract: We challenge text-to-image models with generating escape room puzzle images that are visually appealing, logically solid, and intellectually stimulating. While base image models struggle with spatial relationships and affordance reasoning, we propose a hierarchical multi-agent framework that decomposes this task into structured stages: functional design, symbolic scene graph reasoning, layout synthesis, and local image editing. Specialized agents collaborate through iterative feedback to ensure the scene is visually coherent and functionally solvable. Experiments show that agent collaboration improves output quality in terms of solvability, shortcut avoidance, and affordance clarity, while maintaining visual quality.
Authors: Yuxiang Ge, Jionghao Cheng, Ruiquan Ge, Zhaojie Fang, Gangyong Jia, Xiang Wan, Nannan Li, Ahmed Elazab, Changmiao Wang
Abstract: Reconstructing 3D visual stimuli from Electroencephalography (EEG) data holds significant potential for applications in Brain-Computer Interfaces (BCIs) and aiding individuals with communication disorders. Traditionally, efforts have focused on converting brain activity into 2D images, neglecting the translation of EEG data into 3D objects. This limitation is noteworthy, as the human brain inherently processes three-dimensional spatial information regardless of whether observing 2D images or the real world. The neural activities captured by EEG contain rich spatial information that is inevitably lost when reconstructing only 2D images, thus limiting its practical applications in BCI. The transition from EEG data to 3D object reconstruction faces considerable obstacles. These include the presence of extensive noise within EEG signals and a scarcity of datasets that include both EEG and 3D information, which complicates the extraction process of 3D visual data. Addressing this challenging task, we propose an innovative EEG encoder architecture that integrates a dual self-attention mechanism. We use a hybrid training strategy to train the EEG Encoder, which includes cross-attention, contrastive learning, and self-supervised learning techniques. Additionally, by employing stable diffusion as a prior distribution and utilizing Variational Score Distillation to train a neural radiation field, we successfully generate 3D objects with similar content and structure from EEG data.
Authors: Haofeng Wang, Fangtao Zhou, Qi Zhang, Zeyuan Chen, Enci Zhang, Zhao Wang, Xiaofeng Huang, Siwei Ma
Abstract: RGB-IR(RGB-Infrared) image pairs are frequently applied simultaneously in various applications like intelligent surveillance. However, as the number of modalities increases, the required data storage and transmission costs also double. Therefore, efficient RGB-IR data compression is essential. This work proposes a joint compression framework for RGB-IR image pair. Specifically, to fully utilize cross-modality prior information for accurate context probability modeling within and between modalities, we propose a Channel-wise Cross-modality Entropy Model (CCEM). Among CCEM, a Low-frequency Context Extraction Block (LCEB) and a Low-frequency Context Fusion Block (LCFB) are designed for extracting and aggregating the global low-frequency information from both modalities, which assist the model in predicting entropy parameters more accurately. Experimental results demonstrate that our approach outperforms existing RGB-IR image pair and single-modality compression methods on LLVIP and KAIST datasets. For instance, the proposed framework achieves a 23.1% bit rate saving on LLVIP dataset compared to the state-of-the-art RGB-IR image codec presented at CVPR 2022.
Authors: Jiho Choi, Sang Jun Lee
Abstract: In this paper, we propose a method that learns a general representation of periodic signals from unlabeled facial videos by capturing subtle changes in skin tone over time. The proposed framework employs the video masked autoencoder to learn a high-dimensional spatio-temporal representation of the facial region through self-supervised learning. Capturing quasi-periodic signals in the video is crucial for remote photoplethysmography (rPPG) estimation. To account for signal periodicity, we apply frame masking in terms of video sampling, which allows the model to capture resampled quasi-periodic signals during the pre-training stage. Moreover, the framework incorporates physiological bandlimit constraints, leveraging the property that physiological signals are sparse within their frequency bandwidth to provide pulse cues to the model. The pre-trained encoder is then transferred to the rPPG task, where it is used to extract physiological signals from facial videos. We evaluate the proposed method through extensive experiments on the PURE, UBFC-rPPG, MMPD, and V4V datasets. Our results demonstrate significant performance improvements, particularly in challenging cross-dataset evaluations. Our code is available at https://github.com/ziiho08/Periodic-MAE.
Authors: Ekaterina Redekop, Mara Pleasure, Zichen Wang, Kimberly Flores, Anthony Sisk, William Speier, Corey W. Arnold
Abstract: The rapid growth of digital pathology and advances in self-supervised deep learning have enabled the development of foundational models for various pathology tasks across diverse diseases. While multimodal approaches integrating diverse data sources have emerged, a critical gap remains in the comprehensive integration of whole-slide images (WSIs) with spatial transcriptomics (ST), which is crucial for capturing critical molecular heterogeneity beyond standard hematoxylin & eosin (H&E) staining. We introduce SPADE, a foundation model that integrates histopathology with ST data to guide image representation learning within a unified framework, in effect creating an ST-informed latent space. SPADE leverages a mixture-of-data experts technique, where experts, created via two-stage feature-space clustering, use contrastive learning to learn representations of co-registered WSI patches and gene expression profiles. Pre-trained on the comprehensive HEST-1k dataset, SPADE is evaluated on 14 downstream tasks, demonstrating significantly superior few-shot performance compared to baseline models, highlighting the benefits of integrating morphological and molecular information into one latent space.
Authors: Boyuan Sun, Jiaxing Zhao, Xihan Wei, Qibin Hou
Abstract: In this paper, we present LLaVA-Scissor, a training-free token compression strategy designed for video multimodal large language models. Previous methods mostly attempt to compress tokens based on attention scores, but fail to effectively capture all semantic regions and often lead to token redundancy. Differently, we propose to leverage the Semantic Connected Components (SCC) approach that assigns tokens to distinct semantic regions within the token set, ensuring comprehensive semantic coverage. The outcome is a two-step spatio-temporal token compression strategy that utilizes SCC in both spatial and temporal domains. This strategy can effectively compress tokens by representing the entire video with a set of non-overlapping semantic tokens. We conduct extensive evaluations of the token compression capabilities of LLaVA-Scissor across diverse video understanding benchmarks, including video question answering, long video understanding, and comprehensive multi-choices benchmarks. Experimental results show that the proposed LLaVA-Scissor outperforms other token compression methods, achieving superior performance in various video understanding benchmarks, particularly at low token retention ratios. Project page: https://github.com/HumanMLLM/LLaVA-Scissor.
Authors: Sungjune Park, Yeongyun Kim, Se Yeon Kim, Yong Man Ro
Abstract: Large Vision and Language Models (LVLMs) have shown strong performance across various vision-language tasks in natural image domains. However, their application to remote sensing (RS) remains underexplored due to significant domain differences in visual appearances, object scales, and semantics. These discrepancies hider the effective understanding of RS scenes, which contain rich, multi-level semantic information spanning from coarse-to-fine levels. Hence, it limits the direct adaptation of existing LVLMs to RS imagery. To address this gap, we propose a novel LVLM framework tailored for RS understanding, incorporating two core components: Semantic-augmented Multi-level Alignment and Semantic-aware Expert Modeling. First, to align multi-level visual features, we introduce the retrieval-based Semantic Augmentation Module which enriches the visual features with relevant semantics across fine-to-coarse levels (e.g., object- and scene-level information). It is designed to retrieve relevant semantic cues from a RS semantic knowledge database, followed by aggregation of semantic cues with user query and multi-level visual features, resulting in semantically enriched representation across multiple levels. Second, for Semantic-aware Expert Modeling, we design semantic experts, where each expert is responsible for processing semantic representation at different levels separately. This enables hierarchical semantic understanding from coarse to fine levels. Evaluations across multiple RS tasks-including scene classification and VQA, etc.-demonstrate that the proposed framework achieves consistent improvements across multiple semantic levels. This highlights its capability and effectiveness in bridging the gap between general LVLMs and unique demands of RS-specific vision-language understanding.
Authors: Yanguang Sun, Jiexi Yan, Jianjun Qian, Chunyan Xu, Jian Yang, Lei Luo
Abstract: Automatically segmenting objects from optical remote sensing images (ORSIs) is an important task. Most existing models are primarily based on either convolutional or Transformer features, each offering distinct advantages. Exploiting both advantages is valuable research, but it presents several challenges, including the heterogeneity between the two types of features, high complexity, and large parameters of the model. However, these issues are often overlooked in existing the ORSIs methods, causing sub-optimal segmentation. For that, we propose a novel Dual-Perspective United Transformer (DPU-Former) with a unique structure designed to simultaneously integrate long-range dependencies and spatial details. In particular, we design the global-local mixed attention, which captures diverse information through two perspectives and introduces a Fourier-space merging strategy to obviate deviations for efficient fusion. Furthermore, we present a gated linear feed-forward network to increase the expressive ability. Additionally, we construct a DPU-Former decoder to aggregate and strength features at different layers. Consequently, the DPU-Former model outperforms the state-of-the-art methods on multiple datasets. Code: https://github.com/CSYSI/DPU-Former.
Authors: Tzu-Chun Chien, Chieh-Kai Lin, Shiang-Feng Tsai, Ruei-Chi Lai, Hung-Jen Chen, Min Sun
Abstract: Recent Multimodal Large Language Models (MLLMs) have demonstrated strong performance in visual grounding, establishing themselves as a general interface for various vision-language applications. This progress has driven the development of token pruning methods to mitigate the high computational costs associated with processing numerous visual tokens. However, we observe that pruning significantly weakens the model's grounding ability, leading to incorrect predictions and drastic performance degradation. In Referring Expression Comprehension (REC), for instance, pruning causes the accuracy of LLaVA on the RefCOCO validation set to drop from 56.14% to 15.34%. Our analysis identifies misaligned position IDs after pruning as the primary cause of this degradation, as both the order and value of these IDs are crucial for maintaining performance in grounding tasks. To address this issue, we propose Grounding-Aware Token Pruning (GAP), a simple yet effective adjustment to position IDs that recovers REC accuracy back to 51.42%, which is 90% of the original performance in the without pruning setting, all while requiring no additional training, memory, or computational overhead. Applied to models such as Shikra, MiniGPTv2, and the LLaVA series, our method consistently improves performance across various token pruning strategies.
Authors: Basudha Pal, Sharif Amit Kamran, Brendon Lutnick, Molly Lucas, Chaitanya Parmar, Asha Patel Shah, David Apfel, Steven Fakharzadeh, Lloyd Miller, Gabriela Cula, Kristopher Standish
Abstract: Psoriasis (PsO) severity scoring is important for clinical trials but is hindered by inter-rater variability and the burden of in person clinical evaluation. Remote imaging using patient captured mobile photos offers scalability but introduces challenges, such as variation in lighting, background, and device quality that are often imperceptible to humans but can impact model performance. These factors, along with inconsistencies in dermatologist annotations, reduce the reliability of automated severity scoring. We propose a framework to automatically flag problematic training images that introduce spurious correlations which degrade model generalization, using a gradient based interpretability approach. By tracing the gradients of misclassified validation images, we detect training samples where model errors align with inconsistently rated examples or are affected by subtle, nonclinical artifacts. We apply this method to a ConvNeXT based weakly supervised model designed to classify PsO severity from phone images. Removing 8.2% of flagged images improves model AUC-ROC by 5% (85% to 90%) on a held out test set. Commonly, multiple annotators and an adjudication process ensure annotation accuracy, which is expensive and time consuming. Our method detects training images with annotation inconsistencies, potentially removing the need for manual review. When applied to a subset of training data rated by two dermatologists, the method identifies over 90% of cases with inter-rater disagreement by reviewing only the top 30% of samples. This improves automated scoring for remote assessments, ensuring robustness despite data collection variability.
Authors: Chuheng Wei, Ziye Qin, Ziyan Zhang, Guoyuan Wu, Matthew J. Barth
Abstract: Multi-sensor fusion plays a critical role in enhancing perception for autonomous driving, overcoming individual sensor limitations, and enabling comprehensive environmental understanding. This paper first formalizes multi-sensor fusion strategies into data-level, feature-level, and decision-level categories and then provides a systematic review of deep learning-based methods corresponding to each strategy. We present key multi-modal datasets and discuss their applicability in addressing real-world challenges, particularly in adverse weather conditions and complex urban environments. Additionally, we explore emerging trends, including the integration of Vision-Language Models (VLMs), Large Language Models (LLMs), and the role of sensor fusion in end-to-end autonomous driving, highlighting its potential to enhance system adaptability and robustness. Our work offers valuable insights into current methods and future directions for multi-sensor fusion in autonomous driving.
Authors: Umihiro Kamoto, Tatsuya Ishibashi, Noriyuki Kugo
Abstract: In this report, we present the winning solution that achieved the 1st place in the Complex Video Reasoning & Robustness Evaluation Challenge 2025. This challenge evaluates the ability to generate accurate natural language answers to questions about diverse, real-world video clips. It uses the Complex Video Reasoning and Robustness Evaluation Suite (CVRR-ES) benchmark, which consists of 214 unique videos and 2,400 question-answer pairs spanning 11 categories. Our method, DIVE (Deep-search Iterative Video Exploration), adopts an iterative reasoning approach, in which each input question is semantically decomposed and solved through stepwise reasoning and progressive inference. This enables our system to provide highly accurate and contextually appropriate answers to even the most complex queries. Applied to the CVRR-ES benchmark, our approach achieves 81.44% accuracy on the test set, securing the top position among all participants. This report details our methodology and provides a comprehensive analysis of the experimental results, demonstrating the effectiveness of our iterative reasoning framework in achieving robust video question answering. The code is available at https://github.com/PanasonicConnect/DIVE
Authors: Adam Goodge, Xun Xu, Bryan Hooi, Wee Siong Ng, Jingyi Liao, Yongyi Su, Xulei Yang
Abstract: As point cloud data increases in prevalence in a variety of applications, the ability to detect out-of-distribution (OOD) point cloud objects becomes critical for ensuring model safety and reliability. However, this problem remains under-explored in existing research. Inspired by success in the image domain, we propose to exploit advances in 3D vision-language models (3D VLMs) for OOD detection in point cloud objects. However, a major challenge is that point cloud datasets used to pre-train 3D VLMs are drastically smaller in size and object diversity than their image-based counterparts. Critically, they often contain exclusively computer-designed synthetic objects. This leads to a substantial domain shift when the model is transferred to practical tasks involving real objects scanned from the physical environment. In this paper, our empirical experiments show that synthetic-to-real domain shift significantly degrades the alignment of point cloud with their associated text embeddings in the 3D VLM latent space, hindering downstream performance. To address this, we propose a novel methodology called SODA which improves the detection of OOD point clouds through a neighborhood-based score propagation scheme. SODA is inference-based, requires no additional model training, and achieves state-of-the-art performance over existing approaches across datasets and problem settings.
Authors: Fangling Jiang, Qi Li, Weining Wang, Gang Wang, Bing Liu, Zhenan Sun
Abstract: Recently the emergence of novel presentation attacks has drawn increasing attention to face anti-spoofing. However, existing methods tend to memorize data patterns from the training set, resulting in poor generalization to unknown attack types across different scenarios and limited interpretability. To address these challenges, this paper presents a reinforcement fine-tuning-based face anti-spoofing method that stimulates the capabilities of multimodal large language models to think and learn how to solve the anti-spoofing task itself, rather than relying on the memorization of authenticity patterns. We design verifiable class consistent reward and reasoning consistent reward, and employ a GRPO-based optimization strategy to guide the model in exploring reasoning policies from multiple perspectives to maximize expected rewards. As a result, through iterative trial-and-error learning while retaining only high-reward trajectories, the model distills highly generalizable decision-making rules from the extensive solution space to effectively address cross-domain face anti-spoofing tasks. Extensive experimental results demonstrate that our method achieves state-of-the-art cross-domain generalization performance. It generalizes well to diverse unknown attack types in unseen target domains while providing interpretable reasoning for its authenticity decisions without requiring labor-intensive textual annotations for training.
Authors: Dipayan Biswas, Shishir Shah, Jaspal Subhlok
Abstract: Video is transforming education with online courses and recorded lectures supplementing and replacing classroom teaching. Recent research has focused on enhancing information retrieval for video lectures with advanced navigation, searchability, summarization, as well as question answering chatbots. Visual elements like tables, charts, and illustrations are central to comprehension, retention, and data presentation in lecture videos, yet their full potential for improving access to video content remains underutilized. A major factor is that accurate automatic detection of visual elements in a lecture video is challenging; reasons include i) most visual elements, such as charts, graphs, tables, and illustrations, are artificially created and lack any standard structure, and ii) coherent visual objects may lack clear boundaries and may be composed of connected text and visual components. Despite advancements in deep learning based object detection, current models do not yield satisfactory performance due to the unique nature of visual content in lectures and scarcity of annotated datasets. This paper reports on a transfer learning approach for detecting visual elements in lecture video frames. A suite of state of the art object detection models were evaluated for their performance on lecture video datasets. YOLO emerged as the most promising model for this task. Subsequently YOLO was optimized for lecture video object detection with training on multiple benchmark datasets and deploying a semi-supervised auto labeling strategy. Results evaluate the success of this approach, also in developing a general solution to the problem of object detection in lecture videos. Paper contributions include a publicly released benchmark of annotated lecture video frames, along with the source code to facilitate future research.
Authors: Mingquan Liu
Abstract: Fine Grained Visual Categorization (FGVC) remains a challenging task in computer vision due to subtle inter class differences and fragile feature representations. Existing methods struggle in fine grained scenarios, especially when labeled data is scarce. We propose a semi supervised method combining Mamba based feature modeling, region attention, and Bayesian uncertainty. Our approach enhances local to global feature modeling while focusing on key areas during learning. Bayesian inference selects high quality pseudo labels for stability. Experiments show strong performance on FGVC benchmarks with occlusions, demonstrating robustness when labeled data is limited. Code is available at https://github.com/wxqnl/RAUM Net.
Authors: Justin Reinman, Sunwoong Choi
Abstract: CERBERUS is a synthetic benchmark designed to help train and evaluate AI models for detecting cracks and other defects in infrastructure. It includes a crack image generator and realistic 3D inspection scenarios built in Unity. The benchmark features two types of setups: a simple Fly-By wall inspection and a more complex Underpass scene with lighting and geometry challenges. We tested a popular object detection model (YOLO) using different combinations of synthetic and real crack data. Results show that combining synthetic and real data improves performance on real-world images. CERBERUS provides a flexible, repeatable way to test defect detection systems and supports future research in automated infrastructure inspection. CERBERUS is publicly available at https://github.com/justinreinman/Cerberus-Defect-Generator.
URLs: https://github.com/justinreinman/Cerberus-Defect-Generator.
Authors: Xinghan Wang, Kun Xu, Fei Li, Cao Sheng, Jiazhong Yu, Yadong Mu
Abstract: Text-driven human motion generation has recently attracted considerable attention, allowing models to generate human motions based on textual descriptions. However, current methods neglect the influence of human attributes (such as age, gender, weight, and height) which are key factors shaping human motion patterns. This work represents a pilot exploration for bridging this gap. We conceptualize each motion as comprising both attribute information and action semantics, where textual descriptions align exclusively with action semantics. To achieve this, a new framework inspired by Structural Causal Models is proposed to decouple action semantics from human attributes, enabling text-to-semantics prediction and attribute-controlled generation. The resulting model is capable of generating realistic, attribute-aware motion aligned with the user's text and attribute inputs. For evaluation, we introduce HumanAttr, a comprehensive dataset containing attribute annotations for text-motion pairs, setting the first benchmark for attribute-aware text-to-motion generation. Extensive experiments on the new dataset validate our model's effectiveness.
Authors: Nam Quan Nguyen, Xuan Phong Pham, Tuan-Anh Tran
Abstract: The automated reconstruction of the logical arrangement of tables from image data, termed Table Structure Recognition (TSR), is fundamental for semantic data extraction. Recently, researchers have explored a wide range of techniques to tackle this problem, demonstrating significant progress. Each table is a set of vertical and horizontal separators. Following this realization, we present SepFormer, which integrates the split-and-merge paradigm into a single step through separator regression with a DETR-style architecture, improving speed and robustness. SepFormer is a coarse-to-fine approach that predicts table separators from single-line to line-strip separators with a stack of two transformer decoders. In the coarse-grained stage, the model learns to gradually refine single-line segments through decoder layers with additional angle loss. At the end of the fine-grained stage, the model predicts line-strip separators by refining sampled points from each single-line segment. Our SepFormer can run on average at 25.6 FPS while achieving comparable performance with state-of-the-art methods on several benchmark datasets, including SciTSR, PubTabNet, WTW, and iFLYTAB.
Authors: Juming Xiong, Ruining Deng, Jialin Yue, Siqi Lu, Junlin Guo, Marilyn Lionts, Tianyuan Yao, Can Cui, Junchao Zhu, Chongyu Qu, Mengmeng Yin, Haichun Yang, Yuankai Huo
Abstract: Histological analysis plays a crucial role in understanding tissue structure and pathology. While recent advancements in registration methods have improved 2D histological analysis, they often struggle to preserve critical 3D spatial relationships, limiting their utility in both clinical and research applications. Specifically, constructing accurate 3D models from 2D slices remains challenging due to tissue deformation, sectioning artifacts, variability in imaging techniques, and inconsistent illumination. Deep learning-based registration methods have demonstrated improved performance but suffer from limited generalizability and require large-scale training data. In contrast, non-deep-learning approaches offer better generalizability but often compromise on accuracy. In this study, we introduced ZeroReg3D, a novel zero-shot registration pipeline tailored for accurate 3D reconstruction from serial histological sections. By combining zero-shot deep learning-based keypoint matching with optimization-based affine and non-rigid registration techniques, ZeroReg3D effectively addresses critical challenges such as tissue deformation, sectioning artifacts, staining variability, and inconsistent illumination without requiring retraining or fine-tuning. The code has been made publicly available at https://github.com/hrlblab/ZeroReg3D
Authors: Zhao Jin, Rong-Cheng Tu, Jingyi Liao, Wenhao Sun, Xiao Luo, Shunyu Liu, Dacheng Tao
Abstract: 3D Visual Grounding (3DVG) aims to localize target objects within a 3D scene based on natural language queries. To alleviate the reliance on costly 3D training data, recent studies have explored zero-shot 3DVG by leveraging the extensive knowledge and powerful reasoning capabilities of pre-trained LLMs and VLMs. However, existing paradigms tend to emphasize either spatial (3D-based) or semantic (2D-based) understanding, limiting their effectiveness in complex real-world applications. In this work, we introduce SPAZER - a VLM-driven agent that combines both modalities in a progressive reasoning framework. It first holistically analyzes the scene and produces a 3D rendering from the optimal viewpoint. Based on this, anchor-guided candidate screening is conducted to perform a coarse-level localization of potential objects. Furthermore, leveraging retrieved relevant 2D camera images, 3D-2D joint decision-making is efficiently performed to determine the best-matching object. By bridging spatial and semantic reasoning neural streams, SPAZER achieves robust zero-shot grounding without training on 3D-labeled data. Extensive experiments on ScanRefer and Nr3D benchmarks demonstrate that SPAZER significantly outperforms previous state-of-the-art zero-shot methods, achieving notable gains of 9.0% and 10.9% in accuracy.
Authors: Liu Yang, Huiyu Duan, Jiarui Wang, Jing Liu, Menghan Hu, Xiongkuo Min, Guangtao Zhai, Patrick Le Callet
Abstract: With the rapid advancement of Artificial Intelligence Generated Content (AIGC) techniques, AI generated images (AIGIs) have attracted widespread attention, among which AI generated omnidirectional images (AIGODIs) hold significant potential for Virtual Reality (VR) and Augmented Reality (AR) applications. AI generated omnidirectional images exhibit unique quality issues, however, research on the quality assessment and optimization of AI-generated omnidirectional images is still lacking. To this end, this work first studies the quality assessment and distortion-aware saliency prediction problems for AIGODIs, and further presents a corresponding optimization process. Specifically, we first establish a comprehensive database to reflect human feedback for AI-generated omnidirectionals, termed OHF2024, which includes both subjective quality ratings evaluated from three perspectives and distortion-aware salient regions. Based on the constructed OHF2024 database, we propose two models with shared encoders based on the BLIP-2 model to evaluate the human visual experience and predict distortion-aware saliency for AI-generated omnidirectional images, which are named as BLIP2OIQA and BLIP2OISal, respectively. Finally, based on the proposed models, we present an automatic optimization process that utilizes the predicted visual experience scores and distortion regions to further enhance the visual quality of an AI-generated omnidirectional image. Extensive experiments show that our BLIP2OIQA model and BLIP2OISal model achieve state-of-the-art (SOTA) results in the human visual experience evaluation task and the distortion-aware saliency prediction task for AI generated omnidirectional images, and can be effectively used in the optimization process. The database and codes will be released on https://github.com/IntMeGroup/AIGCOIQA to facilitate future research.
Authors: Naftaly Wambugu, Ruisheng Wang, Bo Guo, Tianshu Yu, Sheng Xu, Mohammed Elhassan
Abstract: Land cover maps generated from semantic segmentation of high-resolution remotely sensed images have drawn mucon in the photogrammetry and remote sensing research community. Currently, massive fine-resolution remotely sensed (FRRS) images acquired by improving sensing and imaging technologies become available. However, accurate semantic segmentation of such FRRS images is greatly affected by substantial class disparities, the invisibility of key ground objects due to occlusion, and object size variation. Despite the extraordinary potential in deep convolutional neural networks (DCNNs) in image feature learning and representation, extracting sufficient features from FRRS images for accurate semantic segmentation is still challenging. These challenges demand the deep learning models to learn robust features and generate sufficient feature descriptors. Specifically, learning multi-contextual features to guarantee adequate coverage of varied object sizes from the ground scene and harnessing global-local contexts to overcome class disparities challenge even profound networks. Deeper networks significantly lose spatial details due to gradual downsampling processes resulting in poor segmentation results and coarse boundaries. This article presents a stacked deep residual network (SDRNet) for semantic segmentation from FRRS images. The proposed framework utilizes two stacked encoder-decoder networks to harness long-range semantics yet preserve spatial information and dilated residual blocks (DRB) between each encoder and decoder network to capture sufficient global dependencies thus improving segmentation performance. Our experimental results obtained using the ISPRS Vaihingen and Potsdam datasets demonstrate that the SDRNet performs effectively and competitively against current DCNNs in semantic segmentation.
Authors: Yixin Zha, Chuxin Wang, Wenfei Yang, Tianzhu Zhang
Abstract: Point cloud understanding aims to acquire robust and general feature representations from unlabeled data. Masked point modeling-based methods have recently shown significant performance across various downstream tasks. These pre-training methods rely on random masking strategies to establish the perception of point clouds by restoring corrupted point cloud inputs, which leads to the failure of capturing reasonable semantic relationships by the self-supervised models. To address this issue, we propose Semantic Masked Autoencoder, which comprises two main components: a prototype-based component semantic modeling module and a component semantic-enhanced masking strategy. Specifically, in the component semantic modeling module, we design a component semantic guidance mechanism to direct a set of learnable prototypes in capturing the semantics of different components from objects. Leveraging these prototypes, we develop a component semantic-enhanced masking strategy that addresses the limitations of random masking in effectively covering complete component structures. Furthermore, we introduce a component semantic-enhanced prompt-tuning strategy, which further leverages these prototypes to improve the performance of pre-trained models in downstream tasks. Extensive experiments conducted on datasets such as ScanObjectNN, ModelNet40, and ShapeNetPart demonstrate the effectiveness of our proposed modules.
Authors: Meng Yu, Te Cui, Qitong Chu, Wenjie Song, Yi Yang, Yufeng Yue
Abstract: Reliable semantic segmentation of open environments is essential for intelligent systems, yet significant problems remain: 1) Existing RGB-T semantic segmentation models mainly rely on low-level visual features and lack high-level textual information, which struggle with accurate segmentation when categories share similar visual characteristics. 2) While SAM excels in instance-level segmentation, integrating it with thermal images and text is hindered by modality heterogeneity and computational inefficiency. To address these, we propose TASeg, a text-aware RGB-T segmentation framework by using Low-Rank Adaptation (LoRA) fine-tuning technology to adapt vision foundation models. Specifically, we propose a Dynamic Feature Fusion Module (DFFM) in the image encoder, which effectively merges features from multiple visual modalities while freezing SAM's original transformer blocks. Additionally, we incorporate CLIP-generated text embeddings in the mask decoder to enable semantic alignment, which further rectifies the classification error and improves the semantic understanding accuracy. Experimental results across diverse datasets demonstrate that our method achieves superior performance in challenging scenarios with fewer trainable parameters.
Authors: Biao Wang, Wenwen Li
Abstract: Visual single object tracking aims to continuously localize and estimate the scale of a target in subsequent video frames, given only its initial state in the first frame. This task has traditionally been framed as a template matching problem, evolving through major phases including correlation filters, two-stream networks, and one-stream networks with significant progress achieved. However, these methods typically require explicit classification and regression modeling, depend on supervised training with large-scale datasets, and are limited to the single task of tracking, lacking flexibility. In recent years, multi-modal large language models (MLLMs) have advanced rapidly. Open-source models like Qwen2.5-VL, a flagship MLLMs with strong foundational capabilities, demonstrate excellent performance in grounding tasks. This has spurred interest in applying such models directly to visual tracking. However, experiments reveal that Qwen2.5-VL struggles with template matching between image pairs (i.e., tracking tasks). Inspired by deepseek-R1, we fine-tuned Qwen2.5-VL using the group relative policy optimization (GRPO) reinforcement learning method on a small-scale dataset with a rule-based reward function. The resulting model, R1-Track, achieved notable performance on the GOT-10k benchmark. R1-Track supports flexible initialization via bounding boxes or text descriptions while retaining most of the original model's general capabilities. And we further discuss potential improvements for R1-Track. This rough technical report summarizes our findings as of May 2025.
Authors: Liudi Yang, Yang Bai, George Eskandar, Fengyi Shen, Mohammad Altillawi, Dong Chen, Soumajit Majumder, Ziyuan Liu, Gitta Kutyniok, Abhinav Valada
Abstract: We address the problem of generating long-horizon videos for robotic manipulation tasks. Text-to-video diffusion models have made significant progress in photorealism, language understanding, and motion generation but struggle with long-horizon robotic tasks. Recent works use video diffusion models for high-quality simulation data and predictive rollouts in robot planning. However, these works predict short sequences of the robot achieving one task and employ an autoregressive paradigm to extend to the long horizon, leading to error accumulations in the generated video and in the execution. To overcome these limitations, we propose a novel pipeline that bypasses the need for autoregressive generation. We achieve this through a threefold contribution: 1) we first decompose the high-level goals into smaller atomic tasks and generate keyframes aligned with these instructions. A second diffusion model then interpolates between each of the two generated frames, achieving the long-horizon video. 2) We propose a semantics preserving attention module to maintain consistency between the keyframes. 3) We design a lightweight policy model to regress the robot joint states from generated videos. Our approach achieves state-of-the-art results on two benchmarks in video quality and consistency while outperforming previous policy models on long-horizon tasks.
Authors: Sarthak Ketanbhai Modi, Lim Zi Pong, Shourya Kuchhal, Yoshi Cao, Yupeng Cheng, Teo Yon Shin, Lin Shang-Wei, Zhiming Li
Abstract: The rapid growth in complexity and size of modern deep neural networks (DNNs) has increased challenges related to computational costs and memory usage, spurring a growing interest in efficient model compression techniques. Previous state-of-the-art approach proposes using a Torque-inspired regularization which forces the weights of neural modules around a selected pivot point. Whereas, we observe that the pruning effect of this approach is far from perfect, as the post-trained network is still dense and also suffers from high accuracy drop. In this work, we attribute such ineffectiveness to the default linear force application scheme, which imposes inappropriate force on neural module of different distances. To efficiently prune the redundant and distant modules while retaining those that are close and necessary for effective inference, in this work, we propose Exponential Torque Pruning (ETP), which adopts an exponential force application scheme for regularization. Experimental results on a broad range of domains demonstrate that, though being extremely simple, ETP manages to achieve significantly higher compression rate than the previous state-of-the-art pruning strategies with negligible accuracy drop.
Authors: Zhanyi Lu, Yue Zhou
Abstract: Facial stylization aims to transform facial images into appealing, high-quality stylized portraits, with the critical challenge of accurately learning the target style while maintaining content consistency with the original image. Although previous StyleGAN-based methods have made significant advancements, the generated results still suffer from artifacts or insufficient fidelity to the source image. We argue that these issues stem from neglecting semantic shift of the generator during stylization. Therefore, we propose a facial stylization method that integrates semantic preservation constraint and pseudo-paired supervision to enhance the content correspondence and improve the stylization effect. Additionally, we develop a methodology for creating multi-level pseudo-paired datasets to implement supervisory constraint. Furthermore, building upon our facial stylization framework, we achieve more flexible multimodal and reference-guided stylization without complex network architecture designs or additional training. Experimental results demonstrate that our approach produces high-fidelity, aesthetically pleasing facial style transfer that surpasses previous methods.
Authors: Han Wang, Shengyang Li, Jian Yang, Yuxuan Liu, Yixuan Lv, Zhuang Zhou
Abstract: Detecting and tracking ground objects using earth observation imagery remains a significant challenge in the field of remote sensing. Continuous maritime ship tracking is crucial for applications such as maritime search and rescue, law enforcement, and shipping analysis. However, most current ship tracking methods rely on geostationary satellites or video satellites. The former offer low resolution and are susceptible to weather conditions, while the latter have short filming durations and limited coverage areas, making them less suitable for the real-world requirements of ship tracking. To address these limitations, we present the Hybrid Optical and Synthetic Aperture Radar (SAR) Ship Re-Identification Dataset (HOSS ReID dataset), designed to evaluate the effectiveness of ship tracking using low-Earth orbit constellations of optical and SAR sensors. This approach ensures shorter re-imaging cycles and enables all-weather tracking. HOSS ReID dataset includes images of the same ship captured over extended periods under diverse conditions, using different satellites of different modalities at varying times and angles. Furthermore, we propose a baseline method for cross-modal ship re-identification, TransOSS, which is built on the Vision Transformer architecture. It refines the patch embedding structure to better accommodate cross-modal tasks, incorporates additional embeddings to introduce more reference information, and employs contrastive learning to pre-train on large-scale optical-SAR image pairs, ensuring the model's ability to extract modality-invariant features. Our dataset and baseline method are publicly available on https://github.com/Alioth2000/Hoss-ReID.
Authors: Jialei Chen, Xu Zheng, Danda Pani Paudel, Luc Van Gool, Hiroshi Murase, Daisuke Deguchi
Abstract: Zero-shot Semantic Segmentation (ZSS) aims to segment both seen and unseen classes using supervision from only seen classes. Beyond adaptation-based methods, distillation-based approaches transfer vision-language alignment of vision-language model, e.g., CLIP, to segmentation models. However, such knowledge transfer remains challenging due to: (1) the difficulty of aligning vision-based features with the textual space, which requires combining spatial precision with vision-language alignment; and (2) the semantic gap between CLIP's global representations and the local, fine-grained features of segmentation models. To address challenge (1), we propose Chimera-Seg, which integrates a segmentation backbone as the body and a CLIP-based semantic head as the head, like the Chimera in Greek mythology, combining spatial precision with vision-language alignment. Specifically, Chimera-Seg comprises a trainable segmentation model and a CLIP Semantic Head (CSH), which maps dense features into the CLIP-aligned space. The CSH incorporates a frozen subnetwork and fixed projection layers from the CLIP visual encoder, along with lightweight trainable components. The partial module from CLIP visual encoder, paired with the segmentation model, retains segmentation capability while easing the mapping to CLIP's semantic space. To address challenge (2), we propose Selective Global Distillation (SGD), which distills knowledge from dense features exhibiting high similarity to the CLIP CLS token, while gradually reducing the number of features used for alignment as training progresses. Besides, we also use a Semantic Alignment Module (SAM) to further align dense visual features with semantic embeddings extracted from the frozen CLIP text encoder. Experiments on two benchmarks show improvements of 0.9% and 1.2% in hIoU.
Authors: Hong Nie, Fuyuan Cao, Lu Chen, Fengxin Chen, Yuefeng Zou, Jun Yu
Abstract: Reconstruction and rendering-based talking head synthesis methods achieve high-quality results with strong identity preservation but are limited by their dependence on identity-specific models. Each new identity requires training from scratch, incurring high computational costs and reduced scalability compared to generative model-based approaches. To overcome this limitation, we propose FIAG, a novel 3D speaking head synthesis framework that enables efficient identity-specific adaptation using only a few training footage. FIAG incorporates Global Gaussian Field, which supports the representation of multiple identities within a shared field, and Universal Motion Field, which captures the common motion dynamics across diverse identities. Benefiting from the shared facial structure information encoded in the Global Gaussian Field and the general motion priors learned in the motion field, our framework enables rapid adaptation from canonical identity representations to specific ones with minimal data. Extensive comparative and ablation experiments demonstrate that our method outperforms existing state-of-the-art approaches, validating both the effectiveness and generalizability of the proposed framework. Code is available at: \textit{https://github.com/gme-hong/FIAG}.
Authors: Durgesh K. Singh, Ahcene Boubekki, Qing Cao, Svein Arne Aase, Robert Jenssen, Michael Kampffmeyer
Abstract: Linear measurements of the left ventricle (LV) in the Parasternal Long Axis (PLAX) view using B-mode echocardiography are crucial for cardiac assessment. These involve placing 4-6 landmarks along a virtual scanline (SL) perpendicular to the LV axis near the mitral valve tips. Manual placement is time-consuming and error-prone, while existing deep learning methods often misalign landmarks, causing inaccurate measurements. We propose a novel framework that enhances LV measurement accuracy by enforcing straight-line constraints. A landmark detector is trained on Anatomical M-Mode (AMM) images, computed in real time from B-mode videos, then transformed back to B-mode space. This approach addresses misalignment and reduces measurement errors. Experiments show improved accuracy over standard B-mode methods, and the framework generalizes well across network architectures. Our semi-automatic design includes a human-in-the-loop step where the user only places the SL, simplifying interaction while preserving alignment flexibility and clinical relevance.
Authors: Dechao Meng, Steven Xiao, Xindi Zhang, Guangyuan Wang, Peng Zhang, Qi Wang, Bang Zhang, Liefeng Bo
Abstract: Audio-driven portrait animation, which synthesizes realistic videos from reference images using audio signals, faces significant challenges in real-time generation of high-fidelity, temporally coherent animations. While recent diffusion-based methods improve generation quality by integrating audio into denoising processes, their reliance on frame-by-frame UNet architectures introduces prohibitive latency and struggles with temporal consistency. This paper introduces MirrorMe, a real-time, controllable framework built on the LTX video model, a diffusion transformer that compresses video spatially and temporally for efficient latent space denoising. To address LTX's trade-offs between compression and semantic fidelity, we propose three innovations: 1. A reference identity injection mechanism via VAE-encoded image concatenation and self-attention, ensuring identity consistency; 2. A causal audio encoder and adapter tailored to LTX's temporal structure, enabling precise audio-expression synchronization; and 3. A progressive training strategy combining close-up facial training, half-body synthesis with facial masking, and hand pose integration for enhanced gesture control. Extensive experiments on the EMTD Benchmark demonstrate MirrorMe's state-of-the-art performance in fidelity, lip-sync accuracy, and temporal stability.
Authors: Petr Hruby, Marc Pollefeys
Abstract: We propose a novel approach for estimating the relative pose between rolling shutter cameras using the intersections of line projections with a single scanline per image. This allows pose estimation without explicitly modeling camera motion. Alternatively, scanlines can be selected within a single image, enabling single-view relative pose estimation for scanlines of rolling shutter cameras. Our approach is designed as a foundational building block for rolling shutter structure-from-motion (SfM), where no motion model is required, and each scanline's pose can be computed independently. % We classify minimal solvers for this problem in both generic and specialized settings, including cases with parallel lines and known gravity direction, assuming known intrinsics and no lens distortion. Furthermore, we develop minimal solvers for the parallel-lines scenario, both with and without gravity priors, by leveraging connections between this problem and the estimation of 2D structure from 1D cameras. % Experiments on rolling shutter images from the Fastec dataset demonstrate the feasibility of our approach for initializing rolling shutter SfM, highlighting its potential for further development. % The code will be made publicly available.
Authors: Shaheer U. Saeed, Yipei Wang, Veeru Kasivisvanathan, Brian R. Davidson, Matthew J. Clarkson, Yipeng Hu, Daniel C. Alexander
Abstract: Reasoning is a hallmark of human intelligence, enabling adaptive decision-making in complex and unfamiliar scenarios. In contrast, machine intelligence remains bound to training data, lacking the ability to dynamically refine solutions at inference time. While some recent advances have explored reasoning in machines, these efforts are largely limited to verbal domains such as mathematical problem-solving, where explicit rules govern step-by-step reasoning. Other critical real-world tasks - including visual perception, spatial reasoning, and radiological diagnosis - require non-verbal reasoning, which remains an open challenge. Here we present a novel learning paradigm that enables machine reasoning in vision by allowing performance improvement with increasing thinking time (inference-time compute), even under conditions where labelled data is very limited. Inspired by dual-process theories of human cognition in psychology, our approach integrates a fast-thinking System I module for familiar tasks, with a slow-thinking System II module that iteratively refines solutions using self-play reinforcement learning. This paradigm mimics human reasoning by proposing, competing over, and refining solutions in data-scarce scenarios. We demonstrate superior performance through extended thinking time, compared not only to large-scale supervised learning but also foundation models and even human experts, in real-world vision tasks. These tasks include computer-vision benchmarks and cancer localisation on medical images across five organs, showcasing transformative potential for non-verbal machine reasoning.
Authors: Pei-Kai Huanga, Ya-Ting Chan, Kuan-Wen Chen, Yen-Chun Chou, Shih-Yu Yang, Chiou-Ting Hsu
Abstract: Many remote Heart Rate (HR) measurement methods focus on estimating remote photoplethysmography (rPPG) signals from video clips lasting around 10 seconds but often overlook the need for HR estimation from ultra-short video clips. In this paper, we aim to accurately measure HR from ultra-short 2-second video clips by specifically addressing two key challenges. First, to overcome the limited number of heartbeat cycles in ultra-short video clips, we propose an effective periodicity-guided rPPG estimation method that enforces consistent periodicity between rPPG signals estimated from ultra-short clips and their much longer ground truth signals. Next, to mitigate estimation inaccuracies due to spectral leakage, we propose including a generator to reconstruct longer rPPG signals from ultra-short ones while preserving their periodic consistency to enable more accurate HR measurement. Extensive experiments on four rPPG estimation benchmark datasets demonstrate that our proposed method not only accurately measures HR from ultra-short video clips but also outperform previous rPPG estimation techniques to achieve state-of-the-art performance.
Authors: Zipei Ma, Junzhe Jiang, Yurui Chen, Li Zhang
Abstract: The realistic reconstruction of street scenes is critical for developing real-world simulators in autonomous driving. Most existing methods rely on object pose annotations, using these poses to reconstruct dynamic objects and move them during the rendering process. This dependence on high-precision object annotations limits large-scale and extensive scene reconstruction. To address this challenge, we propose B\'ezier curve Gaussian splatting (B\'ezierGS), which represents the motion trajectories of dynamic objects using learnable B\'ezier curves. This approach fully leverages the temporal information of dynamic objects and, through learnable curve modeling, automatically corrects pose errors. By introducing additional supervision on dynamic object rendering and inter-curve consistency constraints, we achieve reasonable and accurate separation and reconstruction of scene elements. Extensive experiments on the Waymo Open Dataset and the nuPlan benchmark demonstrate that B\'ezierGS outperforms state-of-the-art alternatives in both dynamic and static scene components reconstruction and novel view synthesis.
Authors: Hyeongji Kim, Stine Hansen, Michael Kampffmeyer
Abstract: Common prototype-based medical image few-shot segmentation (FSS) methods model foreground and background classes using class-specific prototypes. However, given the high variability of the background, a more promising direction is to focus solely on foreground modeling, treating the background as an anomaly -- an approach introduced by ADNet. Yet, ADNet faces three key limitations: dependence on a single prototype per class, a focus on binary classification, and fixed thresholds that fail to adapt to patient and organ variability. To address these shortcomings, we propose the Tied Prototype Model (TPM), a principled reformulation of ADNet with tied prototype locations for foreground and background distributions. Building on its probabilistic foundation, TPM naturally extends to multiple prototypes and multi-class segmentation while effectively separating non-typical background features. Notably, both extensions lead to improved segmentation accuracy. Finally, we leverage naturally occurring class priors to define an ideal target for adaptive thresholds, boosting segmentation performance. Taken together, TPM provides a fresh perspective on prototype-based FSS for medical image segmentation. The code can be found at https://github.com/hjk92g/TPM-FSS.
Authors: Ruthvik Bokkasam, Shankar Gangisetty, A. H. Abdul Hafez, C. V. Jawahar
Abstract: With the rapid advancements in autonomous driving, accurately predicting pedestrian behavior has become essential for ensuring safety in complex and unpredictable traffic conditions. The growing interest in this challenge highlights the need for comprehensive datasets that capture unstructured environments, enabling the development of more robust prediction models to enhance pedestrian safety and vehicle navigation. In this paper, we introduce an Indian driving pedestrian dataset designed to address the complexities of modeling pedestrian behavior in unstructured environments, such as illumination changes, occlusion of pedestrians, unsignalized scene types and vehicle-pedestrian interactions. The dataset provides high-level and detailed low-level comprehensive annotations focused on pedestrians requiring the ego-vehicle's attention. Evaluation of the state-of-the-art intention prediction methods on our dataset shows a significant performance drop of up to $\mathbf{15\%}$, while trajectory prediction methods underperform with an increase of up to $\mathbf{1208}$ MSE, defeating standard pedestrian datasets. Additionally, we present exhaustive quantitative and qualitative analysis of intention and trajectory baselines. We believe that our dataset will open new challenges for the pedestrian behavior research community to build robust models. Project Page: https://cvit.iiit.ac.in/research/projects/cvit-projects/iddped
URLs: https://cvit.iiit.ac.in/research/projects/cvit-projects/iddped
Authors: Antje Alex, Jannis Stoppe
Abstract: Accurate digital twins of industrial assets, such as ships and offshore platforms, rely on the precise reconstruction of complex pipe networks. However, manual modelling of pipes from laser scan data is a time-consuming and labor-intensive process. This paper presents a pipeline for automated pipe reconstruction from incomplete laser scan data. The approach estimates a skeleton curve using Laplacian-based contraction, followed by curve elongation. The skeleton axis is then recentred using a rolling sphere technique combined with 2D circle fitting, and refined with a 3D smoothing step. This enables the determination of pipe properties, including radius, length and orientation, and facilitates the creation of detailed 3D models of complex pipe networks. By automating pipe reconstruction, this approach supports the development of digital twins, allowing for rapid and accurate modeling while reducing costs.
Authors: Zhengyun Cheng, Changhao Wang, Guanwen Zhang, Yi Xu, Wei Zhou, Xiangyang Ji
Abstract: Higher-order tensors are well-suited for representing multi-dimensional data, such as color images and videos. Low-rank tensor representation has become essential in machine learning and computer vision, but existing methods like Tucker decomposition offer flexibility at the expense of interpretability. In contrast, while the CANDECOMP/PARAFAC (CP) decomposition provides a more natural and interpretable tensor structure, obtaining sparse solutions remains challenging. Leveraging the rich properties of CP decomposition, we propose a CP-based low-rank tensor function parameterized by neural networks for implicit neural representation (CP-INR). This approach enables continuous data representation beyond structured grids, fully exploiting the non-linearity of tensor data with theoretical guarantees on excess risk bounds. To achieve a sparse CP decomposition, we introduce a variational form of the Schatten-p quasi-norm and prove its relationship to multilinear rank minimization. For smoothness, we propose a regularization term based on the spectral norm of the Jacobian and Hutchinson's trace estimator. Our proposed smoothness regularization is SVD-free and avoids explicit chain rule derivations. It can serve as an alternative to Total Variation (TV) regularization in image denoising tasks and is naturally applicable to continuous data. Extensive experiments on multi-dimensional data recovery tasks, including image inpainting, denoising, and point cloud upsampling, demonstrate the superiority and versatility of our method compared to state-of-the-art approaches.
Authors: Shaojie Zhang, Jiahui Yang, Jianqin Yin, Zhenbo Luo, Jian Luan
Abstract: Multimodal Large Language Models (MLLMs) have demonstrated significant success in visual understanding tasks. However, challenges persist in adapting these models for video comprehension due to the large volume of data and temporal complexity. Existing Video-LLMs using uniform frame sampling often struggle to capture the query-related crucial spatiotemporal clues of videos effectively. In this paper, we introduce Q-Frame, a novel approach for adaptive frame selection and multi-resolution scaling tailored to the video's content and the specific query. Q-Frame employs a training-free, plug-and-play strategy generated by a text-image matching network like CLIP, utilizing the Gumbel-Max trick for efficient frame selection. Q-Frame allows Video-LLMs to process more frames without exceeding computational limits, thereby preserving critical temporal and spatial information. We demonstrate Q-Frame's effectiveness through extensive experiments on benchmark datasets, including MLVU, LongVideoBench, and Video-MME, illustrating its superiority over existing methods and its applicability across various video understanding tasks.
Authors: Amirmohammad Izadi, Mohammad Ali Banayeeanzade, Fatemeh Askari, Ali Rahimiakbar, Mohammad Mahdi Vahedi, Hosein Hasani, Mahdieh Soleymani Baghshah
Abstract: Despite progress in Vision-Language Models (VLMs), their capacity for visual reasoning is often limited by the \textit{binding problem}: the failure to reliably associate perceptual features with their correct visual referents. This limitation underlies persistent errors in tasks such as counting, visual search, scene description, and spatial relationship understanding. A key factor is that current VLMs process visual features largely in parallel, lacking mechanisms for spatially grounded, serial attention. This paper introduces a simple yet effective intervention: augmenting visual inputs with low-level spatial structures (e.g., horizontal lines) and pairing this with a textual prompt that encourages sequential, spatially-aware parsing. We empirically demonstrate substantial performance improvements across core visual reasoning tasks. Specifically, our method improves GPT-4o visual search accuracy by 25.00%, increases counting accuracy by 26.83%, reduces edit distance error in scene description by 0.32, and enhances performance on spatial relationship tasks by 9.50% on a a 2D synthetic dataset. Furthermore, we find that the visual modification is essential for these gains; purely textual strategies, including Chain-of-Thought prompting, are insufficient and can even degrade performance. Our method enhances binding only with a single-query inference, underscoring the importance of visual input design over purely linguistically-based approaches. These findings suggest that low-level visual structuring is a powerful and underexplored direction for improving compositional visual reasoning and could serve as a general strategy for enhancing VLM performance on spatially grounded tasks.
Authors: Ronald Fecso, Jos\'e Morano, Ursula Schmidt-Erfurth, Hrvoje Bogunovi\'c
Abstract: The rise of imaging techniques such as optical coherence tomography (OCT) and advances in deep learning (DL) have enabled clinicians and researchers to streamline retinal disease staging. A popular DL approach is self-supervised learning (SSL), where models learn from vast amounts of unlabeled data, avoiding costly annotation. SSL has allowed the development of foundation models (FMs), large models that can be used for a variety of downstream tasks. However, existing FMs for OCT, trained solely on image data, lack a comprehensive and robust semantic understanding of images, as evidenced by their downstream performance (especially for complex tasks), and thus require supervised fine-tuning (which may be unfeasible) to better adapt to specific applications and populations. To address this, we propose RetFiner, an SSL vision-language refinement scheme that improves the representations of existing FMs and enables their efficient and direct adaptation to specific populations for improved downstream performance. Our method uses a diverse set of training objectives which take advantage of the rich supervisory signal found in textual data. We tested RetFiner on the retinal FMs RETFound, UrFound, and VisionFM, showing significant improvements in linear probing performance on seven highly diverse OCT classification tasks, with an average increase of 5.8, 3.9, and 2.1 percentage points over their baselines, respectively. Our code and model weights are publicly available at https://github.com/ronnief1/RetFiner.
Authors: Taijin Zhao, Heqian Qiu, Yu Dai, Lanxiao Wang, Fanman Meng, Qingbo Wu, Hongliang Li
Abstract: Few-shot object detection (FSOD) aims to detect objects with limited samples for novel classes, while relying on abundant data for base classes. Existing FSOD approaches, predominantly built on the Faster R-CNN detector, entangle objectness recognition and foreground classification within shared feature spaces. This paradigm inherently establishes class-specific objectness criteria and suffers from unrepresentative novel class samples. To resolve this limitation, we propose a Uniform Orthogonal Feature Space (UOFS) optimization framework. First, UOFS decouples the feature space into two orthogonal components, where magnitude encodes objectness and angle encodes classification. This decoupling enables transferring class-agnostic objectness knowledge from base classes to novel classes. Moreover, implementing the disentanglement requires careful attention to two challenges: (1) Base set images contain unlabeled foreground instances, causing confusion between potential novel class instances and backgrounds. (2) Angular optimization depends exclusively on base class foreground instances, inducing overfitting of angular distributions to base classes. To address these challenges, we propose a Hybrid Background Optimization (HBO) strategy: (1) Constructing a pure background base set by removing unlabeled instances in original images to provide unbiased magnitude-based objectness supervision. (2) Incorporating unlabeled foreground instances in the original base set into angular optimization to enhance distribution uniformity. Additionally, we propose a Spatial-wise Attention Disentanglement and Association (SADA) module to address task conflicts between class-agnostic and class-specific tasks. Experiments demonstrate that our method significantly outperforms existing approaches based on entangled feature spaces.
Authors: Wenhan Wu, Zhishuai Guo, Chen Chen, Hongfei Xue, Aidong Lu
Abstract: Zero-shot skeleton-based action recognition aims to develop models capable of identifying actions beyond the categories encountered during training. Previous approaches have primarily focused on aligning visual and semantic representations but often overlooked the importance of fine-grained action patterns in the semantic space (e.g., the hand movements in drinking water and brushing teeth). To address these limitations, we propose a Frequency-Semantic Enhanced Variational Autoencoder (FS-VAE) to explore the skeleton semantic representation learning with frequency decomposition. FS-VAE consists of three key components: 1) a frequency-based enhancement module with high- and low-frequency adjustments to enrich the skeletal semantics learning and improve the robustness of zero-shot action recognition; 2) a semantic-based action description with multilevel alignment to capture both local details and global correspondence, effectively bridging the semantic gap and compensating for the inherent loss of information in skeleton sequences; 3) a calibrated cross-alignment loss that enables valid skeleton-text pairs to counterbalance ambiguous ones, mitigating discrepancies and ambiguities in skeleton and text features, thereby ensuring robust alignment. Evaluations on the benchmarks demonstrate the effectiveness of our approach, validating that frequency-enhanced semantic features enable robust differentiation of visually and semantically similar action clusters, improving zero-shot action recognition.
Authors: Yuxin Cui, Rui Song, Yibin Li, Max Q. -H. Meng, Zhe Min
Abstract: Robust and accurate 2D/3D registration, which aligns preoperative models with intraoperative images of the same anatomy, is crucial for successful interventional navigation. To mitigate the challenge of a limited field of view in single-image intraoperative scenarios, multi-view 2D/3D registration is required by leveraging multiple intraoperative images. In this paper, we propose a novel multi-view 2D/3D rigid registration approach comprising two stages. In the first stage, a combined loss function is designed, incorporating both the differences between predicted and ground-truth poses and the dissimilarities (e.g., normalized cross-correlation) between simulated and observed intraoperative images. More importantly, additional cross-view training loss terms are introduced for both pose and image losses to explicitly enforce cross-view constraints. In the second stage, test-time optimization is performed to refine the estimated poses from the coarse stage. Our method exploits the mutual constraints of multi-view projection poses to enhance the robustness of the registration process. The proposed framework achieves a mean target registration error (mTRE) of $0.79 \pm 2.17$ mm on six specimens from the DeepFluoro dataset, demonstrating superior performance compared to state-of-the-art registration algorithms.
Authors: Ming Zhao, Pingping Liu, Tongshun Zhang, Zhe Zhang
Abstract: Low-light image enhancement presents two primary challenges: 1) Significant variations in low-light images across different conditions, and 2) Enhancement levels influenced by subjective preferences and user intent. To address these issues, we propose ReF-LLE, a novel personalized low-light image enhancement method that operates in the Fourier frequency domain and incorporates deep reinforcement learning. ReF-LLE is the first to integrate deep reinforcement learning into this domain. During training, a zero-reference image evaluation strategy is introduced to score enhanced images, providing reward signals that guide the model to handle varying degrees of low-light conditions effectively. In the inference phase, ReF-LLE employs a personalized adaptive iterative strategy, guided by the zero-frequency component in the Fourier domain, which represents the overall illumination level. This strategy enables the model to adaptively adjust low-light images to align with the illumination distribution of a user-provided reference image, ensuring personalized enhancement results. Extensive experiments on benchmark datasets demonstrate that ReF-LLE outperforms state-of-the-art methods, achieving superior perceptual quality and adaptability in personalized low-light image enhancement.
Authors: Matthias Tsch\"ope, Vitor Fortes Rey, Sogo Pierre Sanon, Paul Lukowicz, Nikolaos Palaiodimopoulos, Maximilian Kiefer-Emmanouilidis
Abstract: Understanding the impact of small quantum gate perturbations, which are common in quantum digital devices but absent in classical computers, is crucial for identifying potential advantages in quantum machine learning. While these perturbations are typically seen as detrimental to quantum computation, they can actually enhance performance by serving as a natural source of data augmentation. Additionally, they can often be efficiently simulated on classical hardware, enabling quantum-inspired approaches to improve classical machine learning methods. In this paper, we investigate random Bloch sphere rotations, which are fundamental SU(2) transformations, as a simple yet effective quantum-inspired data augmentation technique. Unlike conventional augmentations such as flipping, rotating, or cropping, quantum transformations lack intuitive spatial interpretations, making their application to tasks like image classification less straightforward. While common quantum augmentation methods rely on applying quantum models or trainable quanvolutional layers to classical datasets, we focus on the direct application of small-angle Bloch rotations and their effect on classical data. Using the large-scale ImageNet dataset, we demonstrate that our quantum-inspired augmentation method improves image classification performance, increasing Top-1 accuracy by 3%, Top-5 accuracy by 2.5%, and the F$_1$ score from 8% to 12% compared to standard classical augmentation methods. Finally, we examine the use of stronger unitary augmentations. Although these transformations preserve information in principle, they result in visually unrecognizable images with potential applications for privacy computations. However, we show that our augmentation approach and simple SU(2) transformations do not enhance differential privacy and discuss the implications of this limitation.
Authors: Jiahui Zhang, Yurui Chen, Yueming Xu, Ze Huang, Yanpeng Zhou, Yu-Jie Yuan, Xinyue Cai, Guowei Huang, Xingyue Quan, Hang Xu, Li Zhang
Abstract: Leveraging diverse robotic data for pretraining remains a critical challenge. Existing methods typically model the dataset's action distribution using simple observations as inputs. However, these inputs are often incomplete, resulting in a dispersed conditional action distribution-an issue we refer to as coordinate system chaos and state chaos. This inconsistency significantly hampers pretraining efficiency. To address this, we propose 4D-VLA, a novel approach that effectively integrates 4D information into the input to mitigate these sources of chaos. Our model introduces depth and temporal information into visual features with sequential RGB-D inputs, aligning the coordinate systems of the robot and the scene. This alignment endows the model with strong spatiotemporal reasoning capabilities while minimizing training overhead. Additionally, we introduce memory bank sampling, a frame sampling strategy designed to extract informative frames from historical images, further improving effectiveness and efficiency. Experimental results demonstrate that our pretraining method and architectural components substantially enhance model performance. In both simulated and real-world experiments, our model achieves a significant increase in success rate over OpenVLA. To further assess spatial perception and generalization to novel views, we introduce MV-Bench, a multi-view simulation benchmark. Our model consistently outperforms existing methods, demonstrating stronger spatial understanding and adaptability.
Authors: Yu-Cheng Lin, Yu-Syuan Xu, Hao-Wei Chen, Hsien-Kai Kuo, Chun-Yi Lee
Abstract: Image restoration is a key task in low-level computer vision that aims to reconstruct high-quality images from degraded inputs. The emergence of Vision Mamba, which draws inspiration from the advanced state space model Mamba, marks a significant advancement in this field. Vision Mamba demonstrates excellence in modeling long-range dependencies with linear complexity, a crucial advantage for image restoration tasks. Despite its strengths, Vision Mamba encounters challenges in low-level vision tasks, including computational complexity that scales with the number of scanning sequences and local pixel forgetting. To address these limitations, this study introduces Efficient All-Around Mamba (EAMamba), an enhanced framework that incorporates a Multi-Head Selective Scan Module (MHSSM) with an all-around scanning mechanism. MHSSM efficiently aggregates multiple scanning sequences, which avoids increases in computational complexity and parameter count. The all-around scanning strategy implements multiple patterns to capture holistic information and resolves the local pixel forgetting issue. Our experimental evaluations validate these innovations across several restoration tasks, including super resolution, denoising, deblurring, and dehazing. The results validate that EAMamba achieves a significant 31-89% reduction in FLOPs while maintaining favorable performance compared to existing low-level Vision Mamba methods.
Authors: Filippo Merlo, Ece Takmaz, Wenkai Chen, Albert Gatt
Abstract: Natural scenes provide us with rich contexts for object recognition and reference. In particular, knowing what type of scene one is looking at generates expectations about which objects will occur, and what their spatial configuration should be. Do Vision-Language Models (VLMs) learn to rely on scene contexts in a similar way, when generating references to objects? To address this question, we introduce the \textit{Common Objects Out-of-Context (COOCO)} dataset and test to what extent VLMs rely on scene context to refer to objects under different degrees of scene-object congruency, and different perturbations. Our findings show that models leverage scene context adaptively, depending on both the semantic relatedness between object and scene and the level of noise. In particular, models rely more on context under high target-scene congruence or when objects are degraded. Attention analysis reveals that successful object categorisation involves increased focus on the target in mid-level layers, especially under moderate noise, suggesting that VLMs dynamically balance local and contextual information for reference generation. We make our dataset, code and models available at \href{https://github.com/cs-nlp-uu/scenereg}{https://github.com/cs-nlp-uu/scenereg}.
URLs: https://github.com/cs-nlp-uu/scenereg, https://github.com/cs-nlp-uu/scenereg
Authors: Rui Xu, Yunke Wang, Yong Luo, Bo Du
Abstract: Large Vision-Language Models (LVLMs) encode visual inputs as dense sequences of patch-level tokens to capture fine-grained semantics. These visual tokens often outnumber their textual counterparts by a large margin, leading to substantial computational overhead and limiting the scalability of LVLMs in practice. Previous efforts have explored visual token reduction either prior to or within the large language models (LLM). However, most in-LLM reduction approaches rely on text-conditioned interactions, implicitly assuming that textual tokens can reliably capture the importance of visual tokens. In this work, we revisit this assumption and reveal causal, semantic, and spatial forms of cross-modal misalignment. These misalignments undermine the effectiveness of text-guided visual token reduction. To address this, we introduce VisionDrop, a training-free, visual-only pruning framework that selects informative visual tokens based on intra-modal (visual-to-visual) attention, without relying on textual signals. To further suppress redundancy throughout the model hierarchy, we treat the visual encoder and the LLM as a unified system and design a progressive pruning pipeline. Our method performs dominant token selection and lightweight contextual merging at multiple stages, enabling fine-grained visual information to be retained even under aggressive token budgets. Extensive experiments across diverse benchmarks show that VisionDrop achieves consistent improvements over existing methods, despite requiring no additional training or complex modifications. Its simple yet effective design enables efficient inference while preserving strong performance across tasks.
Authors: Mengqi Zhou, Xipeng Wang, Yuxi Wang, Zhaoxiang Zhang
Abstract: Generating realistic 3D indoor scenes from user inputs remains a challenging problem in computer vision and graphics, requiring careful balance of geometric consistency, spatial relationships, and visual realism. While neural generation methods often produce repetitive elements due to limited global spatial reasoning, procedural approaches can leverage constraints for controllable generation but struggle with multi-constraint scenarios. When constraints become numerous, object collisions frequently occur, forcing the removal of furniture items and compromising layout completeness. To address these limitations, we propose RoomCraft, a multi-stage pipeline that converts real images, sketches, or text descriptions into coherent 3D indoor scenes. Our approach combines a scene generation pipeline with a constraint-driven optimization framework. The pipeline first extracts high-level scene information from user inputs and organizes it into a structured format containing room type, furniture items, and spatial relations. It then constructs a spatial relationship network to represent furniture arrangements and generates an optimized placement sequence using a heuristic-based depth-first search (HDFS) algorithm to ensure layout coherence. To handle complex multi-constraint scenarios, we introduce a unified constraint representation that processes both formal specifications and natural language inputs, enabling flexible constraint-oriented adjustments through a comprehensive action space design. Additionally, we propose a Conflict-Aware Positioning Strategy (CAPS) that dynamically adjusts placement weights to minimize furniture collisions and ensure layout completeness. Extensive experiments demonstrate that RoomCraft significantly outperforms existing methods in generating realistic, semantically coherent, and visually appealing room layouts across diverse input modalities.
Authors: Linhao Zhong, Fan Li, Yi Huang, Jianzhuang Liu, Renjing Pei, Fenglong Song
Abstract: Video outpainting is a challenging task that generates new video content by extending beyond the boundaries of an original input video, requiring both temporal and spatial consistency. Many state-of-the-art methods utilize latent diffusion models with U-Net backbones but still struggle to achieve high quality and adaptability in generated content. Diffusion transformers (DiTs) have emerged as a promising alternative because of their superior performance. We introduce OutDreamer, a DiT-based video outpainting framework comprising two main components: an efficient video control branch and a conditional outpainting branch. The efficient video control branch effectively extracts masked video information, while the conditional outpainting branch generates missing content based on these extracted conditions. Additionally, we propose a mask-driven self-attention layer that dynamically integrates the given mask information, further enhancing the model's adaptability to outpainting tasks. Furthermore, we introduce a latent alignment loss to maintain overall consistency both within and between frames. For long video outpainting, we employ a cross-video-clip refiner to iteratively generate missing content, ensuring temporal consistency across video clips. Extensive evaluations demonstrate that our zero-shot OutDreamer outperforms state-of-the-art zero-shot methods on widely recognized benchmarks.
Authors: Paula Carb\'o Cubero, Alberto Jaenal G\'alvez, Andr\'e Mateus, Jos\'e Ara\'ujo, Patric Jensfelt
Abstract: State-of-the-art methods fail to solve visual localization in scenarios where different devices use different sparse feature extraction algorithms to obtain keypoints and their corresponding descriptors. Translating feature descriptors is enough to enable matching. However, performance is drastically reduced in cross-feature detector cases, because current solutions assume common keypoints. This means that the same detector has to be used, which is rarely the case in practice when different descriptors are used. The low repeatability of keypoints, in addition to non-discriminatory and non-distinctive descriptors, make the identification of true correspondences extremely challenging. We present the first method tackling this problem, which performs feature descriptor augmentation targeting cross-detector feature matching, and then feature translation to a latent space. We show that our method significantly improves image matching and visual localization in the cross-feature scenario and evaluate the proposed method on several benchmarks.
Authors: Luigi Russo, Deodato Tapete, Silvia Liberata Ullo, Paolo Gamba
Abstract: Building damage identification shortly after a disaster is crucial for guiding emergency response and recovery efforts. Although optical satellite imagery is commonly used for disaster mapping, its effectiveness is often hampered by cloud cover or the absence of pre-event acquisitions. To overcome these challenges, we introduce a novel multimodal deep learning (DL) framework for detecting building damage using single-date very high resolution (VHR) Synthetic Aperture Radar (SAR) imagery from the Italian Space Agency (ASI) COSMO SkyMed (CSK) constellation, complemented by auxiliary geospatial data. Our method integrates SAR image patches, OpenStreetMap (OSM) building footprints, digital surface model (DSM) data, and structural and exposure attributes from the Global Earthquake Model (GEM) to improve detection accuracy and contextual interpretation. Unlike existing approaches that depend on pre and post event imagery, our model utilizes only post event data, facilitating rapid deployment in critical scenarios. The framework effectiveness is demonstrated using a new dataset from the 2023 earthquake in Turkey, covering multiple cities with diverse urban settings. Results highlight that incorporating geospatial features significantly enhances detection performance and generalizability to previously unseen areas. By combining SAR imagery with detailed vulnerability and exposure information, our approach provides reliable and rapid building damage assessments without the dependency from available pre-event data. Moreover, the automated and scalable data generation process ensures the framework's applicability across diverse disaster-affected regions, underscoring its potential to support effective disaster management and recovery efforts. Code and data will be made available upon acceptance of the paper.
Authors: Hans Gei{\ss}ner, Christian Rathgeb
Abstract: This paper analyses and addresses the performance gap in the fuzzy vault-based \ac{BCS}. We identify unstable error correction capabilities, which are caused by variable feature set sizes and their influence on similarity thresholds, as a key source of performance degradation. This issue is further compounded by information loss introduced through feature type transformations. To address both problems, we propose a novel feature quantization method based on \it{equal frequent intervals}. This method guarantees fixed feature set sizes and supports training-free adaptation to any number of intervals. The proposed approach significantly reduces the performance gap introduced by template protection. Additionally, it integrates seamlessly with existing systems to minimize the negative effects of feature transformation. Experiments on state-of-the-art face, fingerprint, and iris recognition systems confirm that only minimal performance degradation remains, demonstrating the effectiveness of the method across major biometric modalities.
Authors: Nouf Almesafri, Hector Figueiredo, Miguel Arana-Catania
Abstract: This study investigates the performance of the two most relevant computer vision deep learning architectures, Convolutional Neural Network and Vision Transformer, for event-based cameras. These cameras capture scene changes, unlike traditional frame-based cameras with capture static images, and are particularly suited for dynamic environments such as UAVs and autonomous vehicles. The deep learning models studied in this work are ResNet34 and ViT B16, fine-tuned on the GEN1 event-based dataset. The research evaluates and compares these models under both standard conditions and in the presence of simulated noise. Initial evaluations on the clean GEN1 dataset reveal that ResNet34 and ViT B16 achieve accuracies of 88% and 86%, respectively, with ResNet34 showing a slight advantage in classification accuracy. However, the ViT B16 model demonstrates notable robustness, particularly given its pre-training on a smaller dataset. Although this study focuses on ground-based vehicle classification, the methodologies and findings hold significant promise for adaptation to UAV contexts, including aerial object classification and event-based vision systems for aviation-related tasks.
Authors: Tiankai Chen, Yushu Li, Adam Goodge, Fei Teng, Xulei Yang, Tianrui Li, Xun Xu
Abstract: Out-of-distribution (OOD) detection in 3D point cloud data remains a challenge, particularly in applications where safe and robust perception is critical. While existing OOD detection methods have shown progress for 2D image data, extending these to 3D environments involves unique obstacles. This paper introduces a training-free framework that leverages Vision-Language Models (VLMs) for effective OOD detection in 3D point clouds. By constructing a graph based on class prototypes and testing data, we exploit the data manifold structure to enhancing the effectiveness of VLMs for 3D OOD detection. We propose a novel Graph Score Propagation (GSP) method that incorporates prompt clustering and self-training negative prompting to improve OOD scoring with VLM. Our method is also adaptable to few-shot scenarios, providing options for practical applications. We demonstrate that GSP consistently outperforms state-of-the-art methods across synthetic and real-world datasets 3D point cloud OOD detection.
Authors: Yue Zhang, Jilei Sun, Yunhui Guo, Vibhav Gogate
Abstract: Video Large Multimodal Models (VLMMs) have made impressive strides in understanding video content, but they often struggle with abstract and adaptive reasoning-the ability to revise their interpretations when new information emerges. In reality, conclusions are rarely set in stone; additional context can strengthen or weaken an initial inference. To address this, we introduce Defeasible Video Entailment (DVidE), a new task that challenges models to think like doubters, constantly updating their reasoning based on evolving evidence. In DVidE, given a video premise and a textual hypothesis, models must determine whether a new update strengthens or weakens the hypothesis (classification version) or generate a coherent update that modifies the entailment relationship (generation version). For solving the classification task, we propose the Chain of Counterfactual Thought framework, utilizing counterfactual reasoning, ASR-enhanced video content, and rationale refinement to reduce inference bias. For the generation task, we develop a framework that combines ASR output with a Large Language Model (LLM) to produce coherent, contextually relevant updates aligned with the intended strengthener or weakener goals. Additionally, we introduce a novel benchmark dataset, with strengthener/weakener annotations and an LLM-based evaluation metric specifically designed for assessing generative performance. Experimental results demonstrate significant improvements, highlighting our proposed method in enhancing dynamic reasoning capabilities of VLMMs.
Authors: Shih-Han Chou, Shivam Chandhok, James J. Little, Leonid Sigal
Abstract: Vision-Language Models (VLMs) have achieved impressive performance across a wide range of multimodal tasks, yet they often exhibit inconsistent behavior when faced with semantically equivalent inputs, undermining their reliability and robustness. Recent benchmarks, such as MM-R3, highlight that even state-of-the-art VLMs can produce divergent predictions across semantically equivalent inputs, despite maintaining high average accuracy. Prior work addresses this issue by modifying model architectures or conducting large-scale fine-tuning on curated datasets. In contrast, we propose a simple and effective test-time consistency framework that enhances semantic consistency without supervised re-training. Our method is entirely post-hoc, model-agnostic, and applicable to any VLM with access to its weights. Given a single test point, we enforce consistent predictions via two complementary objectives: (i) a Cross-Entropy Agreement Loss that aligns predictive distributions across semantically equivalent inputs, and (ii) a Pseudo-Label Consistency Loss that draws outputs toward a self-averaged consensus. Our method is plug-and-play and leverages information from a single test input itself to improve consistency. Experiments on the MM-R3 benchmark show that our framework yields substantial gains in consistency across state-of-the-art models, establishing a new direction for inference-time adaptation in multimodal learning.
Authors: Yuhao Liu, Tengfei Wang, Fang Liu, Zhenwei Wang, Rynson W. H. Lau
Abstract: Recent advances in deep generative modeling have unlocked unprecedented opportunities for video synthesis. In real-world applications, however, users often seek tools to faithfully realize their creative editing intentions with precise and consistent control. Despite the progress achieved by existing methods, ensuring fine-grained alignment with user intentions remains an open and challenging problem. In this work, we present Shape-for-Motion, a novel framework that incorporates a 3D proxy for precise and consistent video editing. Shape-for-Motion achieves this by converting the target object in the input video to a time-consistent mesh, i.e., a 3D proxy, allowing edits to be performed directly on the proxy and then inferred back to the video frames. To simplify the editing process, we design a novel Dual-Propagation Strategy that allows users to perform edits on the 3D mesh of a single frame, and the edits are then automatically propagated to the 3D meshes of the other frames. The 3D meshes for different frames are further projected onto the 2D space to produce the edited geometry and texture renderings, which serve as inputs to a decoupled video diffusion model for generating edited results. Our framework supports various precise and physically-consistent manipulations across the video frames, including pose editing, rotation, scaling, translation, texture modification, and object composition. Our approach marks a key step toward high-quality, controllable video editing workflows. Extensive experiments demonstrate the superiority and effectiveness of our approach. Project page: https://shapeformotion.github.io/
Authors: Sadra Safadoust, Fabio Tosi, Fatma G\"uney, Matteo Poggi
Abstract: We introduce WarpRF, a training-free general-purpose framework for quantifying the uncertainty of radiance fields. Built upon the assumption that photometric and geometric consistency should hold among images rendered by an accurate model, WarpRF quantifies its underlying uncertainty from an unseen point of view by leveraging backward warping across viewpoints, projecting reliable renderings to the unseen viewpoint and measuring the consistency with images rendered there. WarpRF is simple and inexpensive, does not require any training, and can be applied to any radiance field implementation for free. WarpRF excels at both uncertainty quantification and downstream tasks, e.g., active view selection and active mapping, outperforming any existing method tailored to specific frameworks.
Authors: Xi Chen, Mingkang Zhu, Shaoteng Liu, Xiaoyang Wu, Xiaogang Xu, Yu Liu, Xiang Bai, Hengshuang Zhao
Abstract: This work explores enabling Chain-of-Thought (CoT) reasoning to link visual cues across multiple images. A straightforward solution is to adapt rule-based reinforcement learning for Vision-Language Models (VLMs). However, such methods typically rely on manually curated question-answer pairs, which can be particularly challenging when dealing with fine grained visual details and complex logic across images. Inspired by self-supervised visual representation learning, we observe that images contain inherent constraints that can serve as supervision. Based on this insight, we construct image triplets comprising two augmented views of the same image and a third, similar but distinct image. During training, the model is prompted to generate a reasoning process to compare these images (i.e., determine same or different). Then we optimize the model with rule-based reinforcement learning. Due to the high visual similarity and the presence of augmentations, the model must attend to subtle visual changes and perform logical reasoning to succeed. Experiments show that, although trained solely on visual comparison tasks, the learned reasoning ability generalizes effectively to a wide range of questions. Without relying on any human-annotated question-answer pairs, our method achieves significant improvements on multi-image reasoning benchmarks and shows strong performance on general vision tasks.
Authors: Hyundong Cho, Spencer Lin, Tejas Srinivasan, Michael Saxon, Deuksin Kwon, Natali T. Chavez, Jonathan May
Abstract: Nonverbal communication (NVC) plays an integral role in human language, but studying NVC in general is challenging because of its broad scope and high variance in interpretation among individuals and cultures. However, mime -- the theatrical technique of suggesting intent using only gesture, expression, and movement -- is a subset of NVC that consists of explicit and embodied actions with much lower human interpretation variance. We argue that a solid understanding of mimed actions is a crucial prerequisite for vision-language models capable of interpreting and commanding more subtle aspects of NVC. Hence, we propose Mime Identification Multimodal Evaluation (MIME), a novel video-based question answering benchmark comprising of 86 mimed actions. Constructed with motion capture data, MIME consists of variations of each action with perturbations applied to the character, background, and viewpoint for evaluating recognition robustness. We find that both open-weight and API-based vision-language models perform significantly worse than humans on MIME, motivating the need for increased research for instilling more robust understanding of human gestures.
Authors: Tinh Nguyen, Minh Khue Phan Tran
Abstract: Sign language recognition is crucial for individuals with hearing impairments to break communication barriers. However, previous approaches have had to choose between efficiency and accuracy. Such as RNNs, LSTMs, and GCNs, had problems with vanishing gradients and high computational costs. Despite improving performance, transformer-based methods were not commonly used. This study presents a new novel SLR approach that overcomes the challenge of independently extracting meaningful information from the x and y coordinates of skeleton sequences, which traditional models often treat as inseparable. By utilizing an encoder-decoder of BART architecture, the model independently encodes the x and y coordinates, while Cross-Attention ensures their interrelation is maintained. With only 749,888 parameters, the model achieves 96.04% accuracy on the LSA-64 dataset, significantly outperforming previous models with over one million parameters. The model also demonstrates excellent performance and generalization across WLASL and ASL-Citizen datasets. Ablation studies underscore the importance of coordinate projection, normalization, and using multiple skeleton components for boosting model efficacy. This study offers a reliable and effective approach for sign language recognition, with strong potential for enhancing accessibility tools for the deaf and hard of hearing.
Authors: Duong Bach
Abstract: Multi-vector document retrieval systems, such as ColPali, excel in fine-grained matching for complex queries but incur significant storage and computational costs due to their reliance on high-dimensional patch embeddings and late-interaction scoring. To address these challenges, we propose HPC-ColPali, a Hierarchical Patch Compression framework that enhances the efficiency of ColPali while preserving its retrieval accuracy. Our approach integrates three innovative techniques: (1) K-Means quantization, which compresses patch embeddings into 1-byte centroid indices, achieving up to 32$\times$ storage reduction; (2) attention-guided dynamic pruning, utilizing Vision-Language Model attention weights to retain only the top-$p\%$ most salient patches, reducing late-interaction computation by up to 60\% with less than 2\% nDCG@10 loss; and (3) optional binary encoding of centroid indices into $b$-bit strings ($b=\lceil\log_2 K\rceil$), enabling rapid Hamming distance-based similarity search for resource-constrained environments. Evaluated on the ViDoRe and SEC-Filings datasets, HPC-ColPali achieves 30--50\% lower query latency under HNSW indexing while maintaining high retrieval precision. When integrated into a Retrieval-Augmented Generation pipeline for legal summarization, it reduces hallucination rates by 30\% and halves end-to-end latency. These advancements establish HPC-ColPali as a scalable and efficient solution for multi-vector document retrieval across diverse applications. Code is available at https://github.com/DngBack/HPC-ColPali.
Authors: Varun Mannam, Fang Wang, Xin Chen
Abstract: Current evaluation frameworks for multimodal generative AI struggle to establish trustworthiness, hindering enterprise adoption where reliability is paramount. We introduce a systematic, quantitative benchmarking framework to measure the trustworthiness of progressively integrating cross-modal inputs such as text, images, captions, and OCR within VisualRAG systems for enterprise document intelligence. Our approach establishes quantitative relationships between technical metrics and user-centric trust measures. Evaluation reveals that optimal modality weighting with weights of 30% text, 15% image, 25% caption, and 30% OCR improves performance by 57.3% over text-only baselines while maintaining computational efficiency. We provide comparative assessments of foundation models, demonstrating their differential impact on trustworthiness in caption generation and OCR extraction-a vital consideration for reliable enterprise AI. This work advances responsible AI deployment by providing a rigorous framework for quantifying and enhancing trustworthiness in multimodal RAG for critical enterprise applications.
Authors: Yixin Sun, Li Li, Wenke E, Amir Atapour-Abarghouei, Toby P. Breckon
Abstract: Detecting traversable pathways in unstructured outdoor environments remains a significant challenge for autonomous robots, especially in critical applications such as wide-area search and rescue, as well as incident management scenarios like forest fires. Existing datasets and models primarily target urban settings or wide, vehicle-traversable off-road tracks, leaving a substantial gap in addressing the complexity of narrow, trail-like off-road scenarios. To address this, we introduce the Trail-based Off-road Multimodal Dataset (TOMD), a comprehensive dataset specifically designed for such environments. TOMD features high-fidelity multimodal sensor data -- including 128-channel LiDAR, stereo imagery, GNSS, IMU, and illumination measurements -- collected through repeated traversals under diverse conditions. We also propose a dynamic multiscale data fusion model for accurate traversable pathway prediction. The study analyzes the performance of early, cross, and mixed fusion strategies under varying illumination levels. Results demonstrate the effectiveness of our approach and the relevance of illumination in segmentation performance. We publicly release TOMD at https://github.com/yyyxs1125/TMOD to support future research in trail-based off-road navigation.
Authors: Haiping Yang, Huaxing Liu, Wei Wu, Zuohui Chen, Ning Wu
Abstract: Unmanned aerial vehicles (UAVs) are increasingly employed in diverse applications such as land surveying, material transport, and environmental monitoring. Following missions like data collection or inspection, UAVs must land safely at docking stations for storage or recharging, which is an essential requirement for ensuring operational continuity. However, accurate landing remains challenging due to factors like GPS signal interference. To address this issue, we propose a deviation warning system for UAV landings, powered by a novel vision-based model called AeroLite-MDNet. This model integrates a multiscale fusion module for robust cross-scale object detection and incorporates a segmentation branch for efficient orientation estimation. We introduce a new evaluation metric, Average Warning Delay (AWD), to quantify the system's sensitivity to landing deviations. Furthermore, we contribute a new dataset, UAVLandData, which captures real-world landing deviation scenarios to support training and evaluation. Experimental results show that our system achieves an AWD of 0.7 seconds with a deviation detection accuracy of 98.6\%, demonstrating its effectiveness in enhancing UAV landing reliability. Code will be available at https://github.com/ITTTTTI/Maskyolo.git
Authors: Minjie Hong, Zirun Guo, Yan Xia, Zehan Wang, Ziang Zhang, Tao Jin, Zhou Zhao
Abstract: Multimodal Large Language Models (MLLMs) are powerful at integrating diverse data, but they often struggle with complex reasoning. While Reinforcement learning (RL) can boost reasoning in LLMs, applying it to MLLMs is tricky. Common issues include a drop in performance on general tasks and the generation of overly detailed or "overthinking" reasoning. Our work investigates how the KL penalty and overthinking affect RL training in MLLMs. We propose Asymmetric Policy Optimization (APO) to address these issues, which divides the sampled responses into positive and negative groups. For positive samples, Difficulty-Adaptive Divergence Shaping (DADS) is introduced to dynamically adjust the KL divergence weight based on their difficulty. This method prevents policy entropy from dropping sharply, improves training stability, utilizes samples better, and preserves the model's existing knowledge. For negative samples, Suboptimal Trajectory Complexity Regularization (STCR) is proposed to penalize overly long responses. This helps mitigate overthinking and encourages more concise reasoning while preserving the model's explorative capacity. We apply our method to Qwen2.5-VL-3B, creating View-R1-3B. View-R1-3B significantly enhances reasoning capabilities, showing an average 7\% gain over the base model and outperforming larger MLLMs (7-11B) on various reasoning benchmarks. Importantly, unlike other reasoning-tuned MLLMs that often degrade on general tasks, View-R1-3B maintains consistent improvement, demonstrating superior generalization. These results highlight the effectiveness and broad applicability of our DADS and STCR techniques for advancing complex multimodal reasoning in MLLMs. The code will be made available at https://github.com/Indolent-Kawhi/View-R1.
Authors: Sai Sri Teja, Sreevidya Chintalapati, Vinayak Gupta, Mukund Varma T, Haejoon Lee, Aswin Sankaranarayanan, Kaushik Mitra
Abstract: Advances in 3D reconstruction using neural rendering have enabled high-quality 3D capture. However, they often fail when the input imagery is corrupted by motion blur, due to fast motion of the camera or the objects in the scene. This work advances neural rendering techniques in such scenarios by using single-photon avalanche diode (SPAD) arrays, an emerging sensing technology capable of sensing images at extremely high speeds. However, the use of SPADs presents its own set of unique challenges in the form of binary images, that are driven by stochastic photon arrivals. To address this, we introduce PhotonSplat, a framework designed to reconstruct 3D scenes directly from SPAD binary images, effectively navigating the noise vs. blur trade-off. Our approach incorporates a novel 3D spatial filtering technique to reduce noise in the renderings. The framework also supports both no-reference using generative priors and reference-based colorization from a single blurry image, enabling downstream applications such as segmentation, object detection and appearance editing tasks. Additionally, we extend our method to incorporate dynamic scene representations, making it suitable for scenes with moving objects. We further contribute PhotonScenes, a real-world multi-view dataset captured with the SPAD sensors.
Authors: Denis Gudovskiy, Wenzhao Zheng, Tomoyuki Okuno, Yohei Nakata, Kurt Keutzer
Abstract: Recently, continuous normalizing flows (CNFs) and diffusion models (DMs) have been studied using the unified theoretical framework. Although such models can generate high-quality data points from a noise distribution, the sampling demands multiple iterations to solve an ordinary differential equation (ODE) with high computational complexity. Most existing methods focus on reducing the number of time steps during the sampling process to improve efficiency. In this work, we explore a complementary direction in which the quality-complexity tradeoff can be dynamically controlled in terms of time steps and in the length of the neural network. We achieve this by rewiring the blocks in the transformer-based architecture to solve an inner discretized ODE w.r.t. its length. Then, we employ time- and length-wise consistency terms during flow matching training, and as a result, the sampling can be performed with an arbitrary number of time steps and transformer blocks. Unlike others, our $\textrm{ODE}_t \left(\textrm{ODE}_l \right)$ approach is solver-agnostic in time dimension and decreases both latency and memory usage. Compared to the previous state of the art, image generation experiments on CelebA-HQ and ImageNet show a latency reduction of up to $3\times$ in the most efficient sampling mode, and a FID score improvement of up to $3.5$ points for high-quality sampling. We release our code and model weights with fully reproducible experiments.
Authors: Ameya Salvi, Venkat Krovi
Abstract: Vision-based lane keeping is a topic of significant interest in the robotics and autonomous ground vehicles communities in various on-road and off-road applications. The skid-steered vehicle architecture has served as a useful vehicle platform for human controlled operations. However, systematic modeling, especially of the skid-slip wheel terrain interactions (primarily in off-road settings) has created bottlenecks for automation deployment. End-to-end learning based methods such as imitation learning and deep reinforcement learning, have gained prominence as a viable deployment option to counter the lack of accurate analytical models. However, the systematic formulation and subsequent verification/validation in dynamic operation regimes (particularly for skid-steered vehicles) remains a work in progress. To this end, a novel approach for structured formulation for learning visual navigation is proposed and investigated in this work. Extensive software simulations, hardware evaluations and ablation studies now highlight the significantly improved performance of the proposed approach against contemporary literature.
Authors: Liav Hen, Erez Yosef, Dan Raviv, Raja Giryes, Jacob Scheuer
Abstract: Metasurfaces are ultra-thin optical elements composed of engineered sub-wavelength structures that enable precise control of light. Their inverse design - determining a geometry that yields a desired optical response - is challenging due to the complex, nonlinear relationship between structure and optical properties. This often requires expert tuning, is prone to local minima, and involves significant computational overhead. In this work, we address these challenges by integrating the generative capabilities of diffusion models into computational design workflows. Using an RCWA simulator, we generate training data consisting of metasurface geometries and their corresponding far-field scattering patterns. We then train a conditional diffusion model to predict meta-atom geometry and height from a target spatial power distribution at a specified wavelength, sampled from a continuous supported band. Once trained, the model can generate metasurfaces with low error, either directly using RCWA-guided posterior sampling or by serving as an initializer for traditional optimization methods. We demonstrate our approach on the design of a spatially uniform intensity splitter and a polarization beam splitter, both produced with low error in under 30 minutes. To support further research in data-driven metasurface design, we publicly release our code and datasets.
Authors: Qi Li, Shaheer U. Saeed, Yuliang Huang, Mingyuan Luo, Zhongnuo Yan, Jiongquan Chen, Xin Yang, Dong Ni, Nektarios Winter, Phuc Nguyen, Lucas Steinberger, Caelan Haney, Yuan Zhao, Mingjie Jiang, Bowen Ren, SiYeoul Lee, Seonho Kim, MinKyung Seo, MinWoo Kim, Yimeng Dou, Zhiwei Zhang, Yin Li, Tomy Varghese, Dean C. Barratt, Matthew J. Clarkson, Tom Vercauteren, Yipeng Hu
Abstract: Trackerless freehand ultrasound reconstruction aims to reconstruct 3D volumes from sequences of 2D ultrasound images without relying on external tracking systems, offering a low-cost, portable, and widely deployable alternative for volumetric imaging. However, it presents significant challenges, including accurate inter-frame motion estimation, minimisation of drift accumulation over long sequences, and generalisability across scanning protocols. The TUS-REC2024 Challenge was established to benchmark and accelerate progress in trackerless 3D ultrasound reconstruction by providing a publicly available dataset for the first time, along with a baseline model and evaluation framework. The Challenge attracted over 43 registered teams, of which 6 teams submitted 21 valid dockerized solutions. Submitted methods spanned a wide range of algorithmic approaches, including recurrent models, registration-driven volume refinement, attention, and physics-informed models. This paper presents an overview of the Challenge design, summarises the key characteristics of the dataset, provides a concise literature review, introduces the technical details of the underlying methodology working with tracked freehand ultrasound data, and offers a comparative analysis of submitted methods across multiple evaluation metrics. The results highlight both the progress and current limitations of state-of-the-art approaches in this domain, and inform directions for future research. The data, evaluation code, and baseline are publicly available to facilitate ongoing development and reproducibility. As a live and evolving benchmark, this Challenge is designed to be continuously developed and improved. The Challenge was held at MICCAI 2024 and will be organised again at MICCAI 2025, reflecting its growing impact and the sustained commitment to advancing this field.
Authors: Avash Palikhe, Zhenyu Yu, Zichong Wang, Wenbin Zhang
Abstract: Large Language Models (LLMs) have played a pivotal role in advancing Artificial Intelligence (AI). However, despite their achievements, LLMs often struggle to explain their decision-making processes, making them a 'black box' and presenting a substantial challenge to explainability. This lack of transparency poses a significant obstacle to the adoption of LLMs in high-stakes domain applications, where interpretability is particularly essential. To overcome these limitations, researchers have developed various explainable artificial intelligence (XAI) methods that provide human-interpretable explanations for LLMs. However, a systematic understanding of these methods remains limited. To address this gap, this survey provides a comprehensive review of explainability techniques by categorizing XAI methods based on the underlying transformer architectures of LLMs: encoder-only, decoder-only, and encoder-decoder models. Then these techniques are examined in terms of their evaluation for assessing explainability, and the survey further explores how these explanations are leveraged in practical applications. Finally, it discusses available resources, ongoing research challenges, and future directions, aiming to guide continued efforts toward developing transparent and responsible LLMs.
Authors: Xiangyu Shi, Yanyuan Qiao, Lingqiao Liu, Feras Dayoub
Abstract: Mobile robots rely on object detectors for perception and object localization in indoor environments. However, standard closed-set methods struggle to handle the diverse objects and dynamic conditions encountered in real homes and labs. Open-vocabulary object detection (OVOD), driven by Vision Language Models (VLMs), extends beyond fixed labels but still struggles with domain shifts in indoor environments. We introduce a Source-Free Domain Adaptation (SFDA) approach that adapts a pre-trained model without accessing source data. We refine pseudo labels via temporal clustering, employ multi-scale threshold fusion, and apply a Mean Teacher framework with contrastive learning. Our Embodied Domain Adaptation for Object Detection (EDAOD) benchmark evaluates adaptation under sequential changes in lighting, layout, and object diversity. Our experiments show significant gains in zero-shot detection performance and flexible adaptation to dynamic indoor conditions.
Authors: Qiyue Gao, Xinyu Pi, Kevin Liu, Junrong Chen, Ruolan Yang, Xinqi Huang, Xinyu Fang, Lu Sun, Gautham Kishore, Bo Ai, Stone Tao, Mengyang Liu, Jiaxi Yang, Chao-Jung Lai, Chuanyang Jin, Jiannan Xiang, Benhao Huang, Zeming Chen, David Danks, Hao Su, Tianmin Shu, Ziqiao Ma, Lianhui Qin, Zhiting Hu
Abstract: Internal world models (WMs) enable agents to understand the world's state and predict transitions, serving as the basis for advanced deliberative reasoning. Recent large Vision-Language Models (VLMs), such as OpenAI o3, GPT-4o and Gemini, exhibit potential as general-purpose WMs. While the latest studies have evaluated and shown limitations in specific capabilities such as visual understanding, a systematic evaluation of VLMs' fundamental WM abilities remains absent. Drawing on comparative psychology and cognitive science, we propose a two-stage framework that assesses Perception (visual, spatial, temporal, quantitative, and motion) and Prediction (mechanistic simulation, transitive inference, compositional inference) to provide an atomic evaluation of VLMs as WMs. Guided by this framework, we introduce WM-ABench, a large-scale benchmark comprising 23 fine-grained evaluation dimensions across 6 diverse simulated environments with controlled counterfactual simulations. Through 660 experiments on 15 latest commercial and open-source VLMs, we find that these models exhibit striking limitations in basic world modeling abilities. For instance, almost all models perform at near-random accuracy when distinguishing motion trajectories. Additionally, they lack disentangled understanding -- e.g., some models tend to believe blue objects move faster than green ones. More rich results and analyses reveal significant gaps between VLMs and human-level world modeling.
Authors: Yuansheng Li, Yunhao Zou, Linwei Chen, Ying Fu
Abstract: Interferometric Hyperspectral Imaging (IHI) is a critical technique for large-scale remote sensing tasks due to its advantages in flux and spectral resolution. However, IHI is susceptible to complex errors arising from imaging steps, and its quality is limited by existing signal processing-based reconstruction algorithms. Two key challenges hinder performance enhancement: 1) the lack of training datasets. 2) the difficulty in eliminating IHI-specific degradation components through learning-based methods. To address these challenges, we propose a novel IHI reconstruction pipeline. First, based on imaging physics and radiometric calibration data, we establish a simplified yet accurate IHI degradation model and a parameter estimation method. This model enables the synthesis of realistic IHI training datasets from hyperspectral images (HSIs), bridging the gap between IHI reconstruction and deep learning. Second, we design the Interferometric Hyperspectral Reconstruction Unfolding Transformer (IHRUT), which achieves effective spectral correction and detail restoration through a stripe-pattern enhancement mechanism and a spatial-spectral transformer architecture. Experimental results demonstrate the superior performance and generalization capability of our method.
Authors: Fabian Perez, Sara Rojas, Carlos Hinojosa, Hoover Rueda-Chac\'on, Bernard Ghanem
Abstract: Neural Radiance Field (NeRF)-based segmentation methods focus on object semantics and rely solely on RGB data, lacking intrinsic material properties. This limitation restricts accurate material perception, which is crucial for robotics, augmented reality, simulation, and other applications. We introduce UnMix-NeRF, a framework that integrates spectral unmixing into NeRF, enabling joint hyperspectral novel view synthesis and unsupervised material segmentation. Our method models spectral reflectance via diffuse and specular components, where a learned dictionary of global endmembers represents pure material signatures, and per-point abundances capture their distribution. For material segmentation, we use spectral signature predictions along learned endmembers, allowing unsupervised material clustering. Additionally, UnMix-NeRF enables scene editing by modifying learned endmember dictionaries for flexible material-based appearance manipulation. Extensive experiments validate our approach, demonstrating superior spectral reconstruction and material segmentation to existing methods. Project page: https://www.factral.co/UnMix-NeRF.
Authors: Najmeh Forouzandehmehr, Reza Yousefi Maragheh, Sriram Kollipara, Kai Zhao, Topojoy Biswas, Evren Korpeoglu, Kannan Achan
Abstract: Automated content-aware layout generation -- the task of arranging visual elements such as text, logos, and underlays on a background canvas -- remains a fundamental yet under-explored problem in intelligent design systems. While recent advances in deep generative models and large language models (LLMs) have shown promise in structured content generation, most existing approaches lack grounding in contextual design exemplars and fall short in handling semantic alignment and visual coherence. In this work we introduce CAL-RAG, a retrieval-augmented, agentic framework for content-aware layout generation that integrates multimodal retrieval, large language models, and collaborative agentic reasoning. Our system retrieves relevant layout examples from a structured knowledge base and invokes an LLM-based layout recommender to propose structured element placements. A vision-language grader agent evaluates the layout with visual metrics, and a feedback agent provides targeted refinements, enabling iterative improvement. We implement our framework using LangGraph and evaluate it on the PKU PosterLayout dataset, a benchmark rich in semantic and structural variability. CAL-RAG achieves state-of-the-art performance across multiple layout metrics -- including underlay effectiveness, element alignment, and overlap -- substantially outperforming strong baselines such as LayoutPrompter. These results demonstrate that combining retrieval augmentation with agentic multi-step reasoning yields a scalable, interpretable, and high-fidelity solution for automated layout generation.
Authors: Shuhan Tan, John Lambert, Hong Jeon, Sakshum Kulshrestha, Yijing Bai, Jing Luo, Dragomir Anguelov, Mingxing Tan, Chiyu Max Jiang
Abstract: The goal of traffic simulation is to augment a potentially limited amount of manually-driven miles that is available for testing and validation, with a much larger amount of simulated synthetic miles. The culmination of this vision would be a generative simulated city, where given a map of the city and an autonomous vehicle (AV) software stack, the simulator can seamlessly simulate the trip from point A to point B by populating the city around the AV and controlling all aspects of the scene, from animating the dynamic agents (e.g., vehicles, pedestrians) to controlling the traffic light states. We refer to this vision as CitySim, which requires an agglomeration of simulation technologies: scene generation to populate the initial scene, agent behavior modeling to animate the scene, occlusion reasoning, dynamic scene generation to seamlessly spawn and remove agents, and environment simulation for factors such as traffic lights. While some key technologies have been separately studied in various works, others such as dynamic scene generation and environment simulation have received less attention in the research community. We propose SceneDiffuser++, the first end-to-end generative world model trained on a single loss function capable of point A-to-B simulation on a city scale integrating all the requirements above. We demonstrate the city-scale traffic simulation capability of SceneDiffuser++ and study its superior realism under long simulation conditions. We evaluate the simulation quality on an augmented version of the Waymo Open Motion Dataset (WOMD) with larger map regions to support trip-level simulation.
Authors: Tianyu Zhang, Xin Luo, Li Li, Dong Liu
Abstract: Diffusion-based image compression has shown remarkable potential for achieving ultra-low bitrate coding (less than 0.05 bits per pixel) with high realism, by leveraging the generative priors of large pre-trained text-to-image diffusion models. However, current approaches require a large number of denoising steps at the decoder to generate realistic results under extreme bitrate constraints, limiting their application in real-time compression scenarios. Additionally, these methods often sacrifice reconstruction fidelity, as diffusion models typically fail to guarantee pixel-level consistency. To address these challenges, we introduce StableCodec, which enables one-step diffusion for high-fidelity and high-realism extreme image compression with improved coding efficiency. To achieve ultra-low bitrates, we first develop an efficient Deep Compression Latent Codec to transmit a noisy latent representation for a single-step denoising process. We then propose a Dual-Branch Coding Structure, consisting of a pair of auxiliary encoder and decoder, to enhance reconstruction fidelity. Furthermore, we adopt end-to-end optimization with joint bitrate and pixel-level constraints. Extensive experiments on the CLIC 2020, DIV2K, and Kodak dataset demonstrate that StableCodec outperforms existing methods in terms of FID, KID and DISTS by a significant margin, even at bitrates as low as 0.005 bits per pixel, while maintaining strong fidelity. Additionally, StableCodec achieves inference speeds comparable to mainstream transform coding schemes. All source code are available at https://github.com/LuizScarlet/StableCodec.
Authors: Qi Gao, Zhihao Chen, Dong Zeng, Junping Zhang, Jianhua Ma, Hongming Shan
Abstract: The generalization of deep learning-based low-dose computed tomography (CT) reconstruction models to doses unseen in the training data is important and remains challenging. Previous efforts heavily rely on paired data to improve the generalization performance and robustness through collecting either diverse CT data for re-training or a few test data for fine-tuning. Recently, diffusion models have shown promising and generalizable performance in low-dose CT (LDCT) reconstruction, however, they may produce unrealistic structures due to the CT image noise deviating from Gaussian distribution and imprecise prior information from the guidance of noisy LDCT images. In this paper, we propose a noise-inspired diffusion model for generalizable LDCT reconstruction, termed NEED, which tailors diffusion models for noise characteristics of each domain. First, we propose a novel shifted Poisson diffusion model to denoise projection data, which aligns the diffusion process with the noise model in pre-log LDCT projections. Second, we devise a doubly guided diffusion model to refine reconstructed images, which leverages LDCT images and initial reconstructions to more accurately locate prior information and enhance reconstruction fidelity. By cascading these two diffusion models for dual-domain reconstruction, our NEED requires only normal-dose data for training and can be effectively extended to various unseen dose levels during testing via a time step matching strategy. Extensive qualitative, quantitative, and segmentation-based evaluations on two datasets demonstrate that our NEED consistently outperforms state-of-the-art methods in reconstruction and generalization performance. Source code is made available at https://github.com/qgao21/NEED.
Authors: Julia Machnio, Sebastian N{\o}rgaard Llambias, Mads Nielsen, Mostafa Mehdipour Ghazi
Abstract: White matter hyperintensities (WMH) are radiological markers of small vessel disease and neurodegeneration, whose accurate segmentation and spatial localization are crucial for diagnosis and monitoring. While multimodal MRI offers complementary contrasts for detecting and contextualizing WM lesions, existing approaches often lack flexibility in handling missing modalities and fail to integrate anatomical localization efficiently. We propose a deep learning framework for WM lesion segmentation and localization that operates directly in native space using single- and multi-modal MRI inputs. Our study evaluates four input configurations: FLAIR-only, T1-only, concatenated FLAIR and T1, and a modality-interchangeable setup. It further introduces a multi-task model for jointly predicting lesion and anatomical region masks to estimate region-wise lesion burden. Experiments conducted on the MICCAI WMH Segmentation Challenge dataset demonstrate that multimodal input significantly improves the segmentation performance, outperforming unimodal models. While the modality-interchangeable setting trades accuracy for robustness, it enables inference in cases with missing modalities. Joint lesion-region segmentation using multi-task learning was less effective than separate models, suggesting representational conflict between tasks. Our findings highlight the utility of multimodal fusion for accurate and robust WMH analysis, and the potential of joint modeling for integrated predictions.
Authors: Noora Sassali, Roel Pieters
Abstract: Pointing gestures are a common interaction method used in Human-Robot Collaboration for various tasks, ranging from selecting targets to guiding industrial processes. This study introduces a method for localizing pointed targets within a planar workspace. The approach employs pose estimation, and a simple geometric model based on shoulder-wrist extension to extract gesturing data from an RGB-D stream. The study proposes a rigorous methodology and comprehensive analysis for evaluating pointing gestures and target selection in typical robotic tasks. In addition to evaluating tool accuracy, the tool is integrated into a proof-of-concept robotic system, which includes object detection, speech transcription, and speech synthesis to demonstrate the integration of multiple modalities in a collaborative application. Finally, a discussion over tool limitations and performance is provided to understand its role in multimodal robotic systems. All developments are available at: https://github.com/NMKsas/gesture_pointer.git.
Authors: Mattia Ricchi, Fabrizio Alfonsi, Camilla Marella, Marco Barbieri, Alessandra Retico, Leonardo Brizi, Alessandro Gabrielli, Claudia Testa
Abstract: Magnetic Resonance Fingerprinting (MRF) is a fast quantitative MR Imaging technique that provides multi-parametric maps with a single acquisition. Neural Networks (NNs) accelerate reconstruction but require significant resources for training. We propose an FPGA-based NN for real-time brain parameter reconstruction from MRF data. Training the NN takes an estimated 200 seconds, significantly faster than standard CPU-based training, which can be up to 250 times slower. This method could enable real-time brain analysis on mobile devices, revolutionizing clinical decision-making and telemedicine.
Authors: Holly Dinkel, Raghavendra Navaratna, Jingyi Xiang, Brian Coltin, Trey Smith, Timothy Bretl
Abstract: This work presents KnotDLO, a method for one-handed Deformable Linear Object (DLO) knot tying that is robust to occlusion, repeatable for varying rope initial configurations, interpretable for generating motion policies, and requires no human demonstrations or training. Grasp and target waypoints for future DLO states are planned from the current DLO shape. Grasp poses are computed from indexing the tracked piecewise linear curve representing the DLO state based on the current curve shape and are piecewise continuous. KnotDLO computes intermediate waypoints from the geometry of the current DLO state and the desired next state. The system decouples visual reasoning from control. In 16 trials of knot tying, KnotDLO achieves a 50% success rate in tying an overhand knot from previously unseen configurations.
Authors: Hao Xu, Ruth Lim, Brian E. Chapman
Abstract: Purpose: Aortic dissections are life-threatening cardiovascular conditions requiring accurate segmentation of true lumen (TL), false lumen (FL), and false lumen thrombosis (FLT) from CTA images for effective management. Manual segmentation is time-consuming and variable, necessitating automated solutions. Materials and Methods: We developed four deep learning-based pipelines for Type B aortic dissection segmentation: a single-step model, a sequential model, a sequential multi-task model, and an ensemble model, utilizing 3D U-Net and Swin-UnetR architectures. A dataset of 100 retrospective CTA images was split into training (n=80), validation (n=10), and testing (n=10). Performance was assessed using the Dice Coefficient and Hausdorff Distance. Results: Our approach achieved superior segmentation accuracy, with Dice Coefficients of 0.91 $\pm$ 0.07 for TL, 0.88 $\pm$ 0.18 for FL, and 0.47 $\pm$ 0.25 for FLT, outperforming Yao et al. (1), who reported 0.78 $\pm$ 0.20, 0.68 $\pm$ 0.18, and 0.25 $\pm$ 0.31, respectively. Conclusion: The proposed pipelines provide accurate segmentation of TBAD features, enabling derivation of morphological parameters for surveillance and treatment planning
Authors: Ajay Mittal, Raghav Mehta, Omar Todd, Philipp Seeb\"ock, Georg Langs, Ben Glocker
Abstract: Automatic detection and classification of Cardiovascular disease (CVD) from Computed Tomography (CT) images play an important part in facilitating better-informed clinical decisions. However, most of the recent deep learning based methods either directly work on raw CT data or utilize it in pair with anatomical cardiac structure segmentation by training an end-to-end classifier. As such, these approaches become much more difficult to interpret from a clinical perspective. To address this challenge, in this work, we break down the CVD classification pipeline into three components: (i) image segmentation, (ii) image registration, and (iii) downstream CVD classification. Specifically, we utilize the Atlas-ISTN framework and recent segmentation foundational models to generate anatomical structure segmentation and a normative healthy atlas. These are further utilized to extract clinically interpretable radiomic features as well as deformation field based geometric features (through atlas registration) for CVD classification. Our experiments on the publicly available ASOCA dataset show that utilizing these features leads to better CVD classification accuracy (87.50\%) when compared against classification model trained directly on raw CT images (67.50\%). Our code is publicly available: https://github.com/biomedia-mira/grc-net
Authors: Yuliang Huang, Imraj Singh, Thomas Joyce, Kris Thielemans, Jamie R. McClelland
Abstract: 3D Cone-Beam CT (CBCT) is widely used in radiotherapy but suffers from motion artifacts due to breathing. A common clinical approach mitigates this by sorting projections into respiratory phases and reconstructing images per phase, but this does not account for breathing variability. Dynamic CBCT instead reconstructs images at each projection, capturing continuous motion without phase sorting. Recent advancements in 4D Gaussian Splatting (4DGS) offer powerful tools for modeling dynamic scenes, yet their application to dynamic CBCT remains underexplored. Existing 4DGS methods, such as HexPlane, use implicit motion representations, which are computationally expensive. While explicit low-rank motion models have been proposed, they lack spatial regularization, leading to inconsistencies in Gaussian motion. To address these limitations, we introduce a free-form deformation (FFD)-based spatial basis function and a deformation-informed framework that enforces consistency by coupling the temporal evolution of Gaussian's mean position, scale, and rotation under a unified deformation field. We evaluate our approach on six CBCT datasets, demonstrating superior image quality with a 6x speedup over HexPlane. These results highlight the potential of deformation-informed 4DGS for efficient, motion-compensated CBCT reconstruction. The code is available at https://github.com/Yuliang-Huang/DIGS.
Authors: Erkan Turan, Aristotelis Siozopoulos, Maks Ovsjanikov
Abstract: Conditional Flow Matching (CFM) offers a simulation-free framework for training continuous-time generative models, bridging diffusion and flow-based approaches. However, sampling from CFM still relies on numerically solving non-linear ODEs which can be computationally expensive and difficult to interpret. Recent alternatives address sampling speed via trajectory straightening, mini-batch coupling or distillation. However, these methods typically do not shed light on the underlying \textit{structure} of the generative process. In this work, we propose to accelerate CFM and introduce an interpretable representation of its dynamics by integrating Koopman operator theory, which models non-linear flows as linear evolution in a learned space of observables. We introduce a decoder-free Koopman-CFM architecture that learns an embedding where the generative dynamics become linear, enabling closed-form, one-step sampling via matrix exponentiation. This results in significant speedups over traditional CFM as demonstrated on controlled 2D datasets and real-world benchmarks, MNIST, Fashion-MNIST (F-MNIST), and the Toronto Face Dataset (TFD). Unlike previous methods, our approach leads to a well-structured Koopman generator, whose spectral properties, eigenvalues, and eigenfunctions offer principled tools for analyzing generative behavior such as temporal scaling, mode stability, and decomposition in Koopman latent space. By combining sampling efficiency with analytical structure, Koopman-enhanced flow matching offers a potential step toward fast and interpretable generative modeling.
Authors: Yannick Werner, Akash Malemath, Mengxi Liu, Vitor Fortes Rey, Nikolaos Palaiodimopoulos, Paul Lukowicz, Maximilian Kiefer-Emmanouilidis
Abstract: Kolmogorov Arnold Networks (KANs), built upon the Kolmogorov Arnold representation theorem (KAR), have demonstrated promising capabilities in expressing complex functions with fewer neurons. This is achieved by implementing learnable parameters on the edges instead of on the nodes, unlike traditional networks such as Multi-Layer Perceptrons (MLPs). However, KANs potential in quantum machine learning has not yet been well explored. In this work, we present an implementation of these KAN architectures in both hybrid and fully quantum forms using a Quantum Circuit Born Machine (QCBM). We adapt the KAN transfer using pre-trained residual functions, thereby exploiting the representational power of parametrized quantum circuits. In the hybrid model we combine classical KAN components with quantum subroutines, while the fully quantum version the entire architecture of the residual function is translated to a quantum model. We demonstrate the feasibility, interpretability and performance of the proposed Quantum KAN (QuKAN) architecture.
Authors: Anirban Ray, Ashesh, Florian Jug
Abstract: Fluorescence microscopy is a major driver of scientific progress in the life sciences. Although high-end confocal microscopes are capable of filtering out-of-focus light, cheaper and more accessible microscopy modalities, such as widefield microscopy, can not, which consequently leads to hazy image data. Computational dehazing is trying to combine the best of both worlds, leading to cheap microscopy but crisp-looking images. The perception-distortion trade-off tells us that we can optimize either for data fidelity, e.g. low MSE or high PSNR, or for data realism, measured by perceptual metrics such as LPIPS or FID. Existing methods either prioritize fidelity at the expense of realism, or produce perceptually convincing results that lack quantitative accuracy. In this work, we propose HazeMatching, a novel iterative method for dehazing light microscopy images, which effectively balances these objectives. Our goal was to find a balanced trade-off between the fidelity of the dehazing results and the realism of individual predictions (samples). We achieve this by adapting the conditional flow matching framework by guiding the generative process with a hazy observation in the conditional velocity field. We evaluate HazeMatching on 5 datasets, covering both synthetic and real data, assessing both distortion and perceptual quality. Our method is compared against 7 baselines, achieving a consistent balance between fidelity and realism on average. Additionally, with calibration analysis, we show that HazeMatching produces well-calibrated predictions. Note that our method does not need an explicit degradation operator to exist, making it easily applicable on real microscopy data. All data used for training and evaluation and our code will be publicly available under a permissive license.
Authors: Xiang Dai, Kyrollos Yanny, Kristina Monakhova, Nicholas Antipa
Abstract: High-dynamic-range (HDR) imaging is an essential technique for overcoming the dynamic range limits of image sensors. The classic method relies on multiple exposures, which slows capture time, resulting in motion artifacts when imaging dynamic scenes. Single-shot HDR imaging alleviates this issue by encoding HDR data into a single exposure, then computationally recovering it. Many established methods use strong image priors to recover improperly exposed image detail. These approaches struggle with extended highlight regions. We utilize the global reset release (GRR) shutter mode of an off-the-shelf sensor. GRR shutter mode applies a longer exposure time to rows closer to the bottom of the sensor. We use optics that relay a randomly permuted (shuffled) image onto the sensor, effectively creating spatially randomized exposures across the scene. The exposure diversity allows us to recover HDR data by solving an optimization problem with a simple total variation image prior. In simulation, we demonstrate that our method outperforms other single-shot methods when many sensor pixels are saturated (10% or more), and is competitive at a modest saturation (1%). Finally, we demonstrate a physical lab prototype that uses an off-the-shelf random fiber bundle for the optical shuffling. The fiber bundle is coupled to a low-cost commercial sensor operating in GRR shutter mode. Our prototype achieves a dynamic range of up to 73dB using an 8-bit sensor with 48dB dynamic range.
Authors: Haosheng Chen, Yue Wu, Yidong Peng
Abstract: Event-based approaches, which are based on bio-inspired asynchronous event cameras, have achieved promising performance on various computer vision tasks. However, the study of the fundamental event data association problem is still in its infancy. In this paper, we propose a novel Event Data Association (called EDA) approach to explicitly address the event association and fusion problem. The proposed EDA seeks for event trajectories that best fit the event data, in order to perform unifying data association and information fusion. In EDA, we first asynchronously fuse the event data based on its information entropy. Then, we introduce a deterministic model hypothesis generation strategy, which effectively generates model hypotheses from the fused events, to represent the corresponding event trajectories. After that, we present a two-stage weighting algorithm, which robustly weighs and selects true models from the generated model hypotheses, through multi-structural geometric model fitting. Meanwhile, we also propose an adaptive model selection strategy to automatically determine the number of the true models. Finally, we use the selected true models to associate and fuse the event data, without being affected by sensor noise and irrelevant structures. We evaluate the performance of the proposed EDA on the object tracking task. The experimental results show the effectiveness of EDA under challenging scenarios, such as high speed, motion blur, and high dynamic range conditions.
Authors: Miko{\l}aj Sacha, Dawid Rymarczyk, {\L}ukasz Struski, Jacek Tabor, Bartosz Zieli\'nski
Abstract: We introduce ProtoSeg, a novel model for interpretable semantic image segmentation, which constructs its predictions using similar patches from the training set. To achieve accuracy comparable to baseline methods, we adapt the mechanism of prototypical parts and introduce a diversity loss function that increases the variety of prototypes within each class. We show that ProtoSeg discovers semantic concepts, in contrast to standard segmentation models. Experiments conducted on Pascal VOC and Cityscapes datasets confirm the precision and transparency of the presented method.
Authors: Junxi Chen, Junhao Dong, Xiaohua Xie, Jianhuang Lai
Abstract: Adversarial training (AT) is considered the most effective defense against adversarial attacks. However, a recent study revealed that \(\ell_{\infty}\)-norm adversarial training (\(\ell_{\infty}\)-AT) will also induce unevenly distributed input gradients, which is called the inequality phenomenon. This phenomenon makes the \(\ell_{\infty}\)-norm adversarially trained model more vulnerable than the standard-trained model when high-attribution or randomly selected pixels are perturbed, enabling robust and practical black-box attacks against \(\ell_{\infty}\)-adversarially trained models. In this paper, we propose a simple yet effective method called Input Gradient Distillation (IGD) to release the inequality phenomenon in $\ell_{\infty}$-AT. IGD distills the standard-trained teacher model's equal decision pattern into the $\ell_{\infty}$-adversarially trained student model by aligning input gradients of the student model and the standard-trained model with the Cosine Similarity. Experiments show that IGD can mitigate the inequality phenomenon and its threats while preserving adversarial robustness. Compared to vanilla $\ell_{\infty}$-AT, IGD reduces error rates against inductive noise, inductive occlusion, random noise, and noisy images in ImageNet-C by up to 60\%, 16\%, 50\%, and 21\%, respectively. Other than empirical experiments, we also conduct a theoretical analysis to explain why releasing the inequality phenomenon can improve such robustness and discuss why the severity of the inequality phenomenon varies according to the dataset's image resolution. Our code is available at https://github.com/fhdnskfbeuv/Inuput-Gradient-Distillation
URLs: https://github.com/fhdnskfbeuv/Inuput-Gradient-Distillation
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
Abstract: 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: Zanting Ye, Ke Wang, Wenbing Lv, Qianjin Feng, Lijun Lu
Abstract: Deep learning-based medical image segmentation faces significant challenges arising from limited labeled data and domain shifts. While prior approaches have primarily addressed these issues independently, their simultaneous occurrence is common in medical imaging. A method that generalizes to unseen domains using only minimal annotations offers significant practical value due to reduced data annotation and development costs. In pursuit of this goal, we propose FSDA-DG, a novel solution to improve cross-domain generalizability of medical image segmentation with few single-source domain annotations. Specifically, our approach introduces semantics-guided semi-supervised data augmentation. This method divides images into global broad regions and semantics-guided local regions, and applies distinct augmentation strategies to enrich data distribution. Within this framework, both labeled and unlabeled data are transformed into extensive domain knowledge while preserving domain-invariant semantic information. Additionally, FSDA-DG employs a multi-decoder U-Net pipeline semi-supervised learning (SSL) network to improve domain-invariant representation learning through consistent prior assumption across multiple perturbations. By integrating data-level and model-level designs, FSDA-DG achieves superior performance compared to state-of-the-art methods in two challenging single domain generalization (SDG) tasks with limited annotations. The code is publicly available at https://github.com/yezanting/FSDA-DG.
Authors: Roie Kazoom, Raz Birman, Ofer Hadar
Abstract: The widespread adoption of computer vision systems has underscored their susceptibility to adversarial attacks, particularly adversarial patch attacks on object detectors. This study evaluates defense mechanisms for the YOLOv5 model against such attacks. Optimized adversarial patches were generated and placed in sensitive image regions, by applying EigenCAM and grid search to determine optimal placement. We tested several defenses, including Segment and Complete (SAC), Inpainting, and Latent Diffusion Models. Our pipeline comprises three main stages: patch application, object detection, and defense analysis. Results indicate that adversarial patches reduce average detection confidence by 22.06\%. Defenses restored confidence levels by 3.45\% (SAC), 5.05\% (Inpainting), and significantly improved them by 26.61\%, which even exceeds the original accuracy levels, when using the Latent Diffusion Model, highlighting its superior effectiveness in mitigating the effects of adversarial patches.
Authors: Timo Kaiser, Maximilian Schier, Bodo Rosenhahn
Abstract: Cell tracking and segmentation assist biologists in extracting insights from large-scale microscopy time-lapse data. Driven by local accuracy metrics, current tracking approaches often suffer from a lack of long-term consistency and the ability to reconstruct lineage trees correctly. To address this issue, we introduce an uncertainty estimation technique for motion estimation frameworks and extend the multi-hypothesis tracking framework. Our uncertainty estimation lifts motion representations into probabilistic spatial densities using problem-specific test-time augmentations. Moreover, we introduce a novel mitosis-aware assignment problem formulation that allows multi-hypothesis trackers to model cell splits and to resolve false associations and mitosis detections based on long-term conflicts. In our framework, explicit biological knowledge is modeled in assignment costs. We evaluate our approach on nine competitive datasets and demonstrate that we outperform the current state-of-the-art on biologically inspired metrics substantially, achieving improvements by a factor of approximately 6 and uncover new insights into the behavior of motion estimation uncertainty.
Authors: Kirillov Alexey, Andrey Moskalenko, Dmitriy Vatolin
Abstract: Over the past few years, deep neural models have made considerable advances in image quality assessment (IQA). However, the underlying reasons for their success remain unclear, owing to the complex nature of deep neural networks. IQA aims to describe how the human visual system (HVS) works and to create its efficient approximations. On the other hand, Saliency Prediction task aims to emulate HVS via determining areas of visual interest. Thus, we believe that saliency plays a crucial role in human perception. In this work, we conduct an empirical study that reveals the relation between IQA and Saliency Prediction tasks, demonstrating that the former incorporates knowledge of the latter. Moreover, we introduce a novel SACID dataset of saliency-aware compressed images and conduct a large-scale comparison of classic and neural-based IQA methods. All supplementary code and data will be available at the time of publication.
Authors: Fangzhou Han, Lingyu Si, Zhizhuo Jiang, Hongwei Dong, Lamei Zhang, Yu Liu, Hao Chen, Bo Du
Abstract: Artificial intelligence generative content (AIGC) has significantly impacted image generation in the field of remote sensing. However, the equally important area of remote sensing image (RSI) editing has not received sufficient attention. Deep learning based editing methods generally involve two sequential stages: generation and editing.For natural images, these stages primarily rely on generative backbones pre-trained on large-scale benchmark datasets and text guidance facilitated by vision-language models (VLMs). However, it become less viable for RSIs: First, existing generative RSI benchmark datasets do not fully capture the diversity of RSIs, and is often inadequate for universal editing tasks. Second, the single text semantic corresponds to multiple image semantics, leading to the introduction of incorrect semantics.To solve above problems, this paper proposes a text-guided RSI editing method and can be trained using only a single image. A multi-scale training approach is adopted to preserve consistency without the need for training on extensive benchmarks, while leveraging RSI pre-trained VLMs and prompt ensembling (PE) to ensure accuracy and controllability. Experimental results on multiple RSI editing tasks show that the proposed method offers significant advantages in both CLIP scores and subjective evaluations compared to existing methods. Additionally, we explore the ability of the edited RSIs to support disaster assessment tasks in order to validate their practicality. Codes will be released at https://github.com/HIT-PhilipHan/remote_sensing_image_editing
URLs: https://github.com/HIT-PhilipHan/remote_sensing_image_editing
Authors: Antonio R\'ios-Vila, Jorge Calvo-Zaragoza, David Rizo, Thierry Paquet
Abstract: Optical Music Recognition (OMR) has made significant progress since its inception, with various approaches now capable of accurately transcribing music scores into digital formats. Despite these advancements, most so-called end-to-end OMR approaches still rely on multi-stage processing pipelines for transcribing full-page score images, which entails challenges such as the need for dedicated layout analysis and specific annotated data, thereby limiting the general applicability of such methods. In this paper, we present the first truly end-to-end approach for page-level OMR in complex layouts. Our system, which combines convolutional layers with autoregressive Transformers, processes an entire music score page and outputs a complete transcription in a music encoding format. This is made possible by both the architecture and the training procedure, which utilizes curriculum learning through incremental synthetic data generation. We evaluate the proposed system using pianoform corpora, which is one of the most complex sources in the OMR literature. This evaluation is conducted first in a controlled scenario with synthetic data, and subsequently against two real-world corpora of varying conditions. Our approach is compared with leading commercial OMR software. The results demonstrate that our system not only successfully transcribes full-page music scores but also outperforms the commercial tool in both zero-shot settings and after fine-tuning with the target domain, representing a significant contribution to the field of OMR.
Authors: Yuang Zhang, Jiaxi Gu, Li-Wen Wang, Han Wang, Junqi Cheng, Yuefeng Zhu, Fangyuan Zou
Abstract: In recent years, generative artificial intelligence has achieved significant advancements in the field of image generation, spawning a variety of applications. However, video generation still faces considerable challenges in various aspects, such as controllability, video length, and richness of details, which hinder the application and popularization of this technology. In this work, we propose a controllable video generation framework, dubbed MimicMotion, which can generate high-quality videos of arbitrary length mimicking specific motion guidance. Compared with previous methods, our approach has several highlights. Firstly, we introduce confidence-aware pose guidance that ensures high frame quality and temporal smoothness. Secondly, we introduce regional loss amplification based on pose confidence, which significantly reduces image distortion. Lastly, for generating long and smooth videos, we propose a progressive latent fusion strategy. By this means, we can produce videos of arbitrary length with acceptable resource consumption. With extensive experiments and user studies, MimicMotion demonstrates significant improvements over previous approaches in various aspects. Detailed results and comparisons are available on our project page: https://tencent.github.io/MimicMotion .
Authors: Xiaojie Li, Yibo Yang, Jianlong Wu, Yue Yu, Ming-Hsuan Yang, Liqiang Nie, Min Zhang
Abstract: Few-shot class-incremental learning (FSCIL) aims to incrementally learn novel classes from limited examples while preserving knowledge of previously learned classes. Existing methods face a critical dilemma: static architectures rely on a fixed parameter space to learn from data that arrive sequentially, prone to overfitting to the current session, while dynamic architectures require the expansion of the parameter space continually, leading to increased complexity. In this study, we explore the potential of Selective State Space Models (SSMs) for FSCIL. Mamba leverages its input-dependent parameters to dynamically adjust its processing patterns and generate content-aware scan patterns within a fixed architecture. This enables it to configure distinct processing for base and novel classes, effectively preserving existing knowledge while adapting to new ones. To leverage Mamba's potential for FSCIL, we design two key modules: First, we propose a dual selective SSM projector that dynamically adjusts the projection parameters based on the intermediate features for dynamic adaptation. The dual-design structurally decouples base and novel class processing with a frozen base branch, employing a frozen base branch to maintain robust base-class features and a dynamic incremental branch that adaptively learns distinctive feature shifts for novel classes. Second, we develop a class-sensitive selective scan mechanism to guide dynamic adaptation of the incremental branch. It minimizes the disruption to base-class representations caused by training on novel data, and meanwhile, forces the selective scan to perform in distinct patterns between base and novel classes. Extensive experiments on miniImageNet, CUB-200, and CIFAR-100 demonstrate that Mamba-FSCIL achieves state-of-the-art performance. The code is available at https://github.com/xiaojieli0903/Mamba-FSCIL.
Authors: Jia-Hau Bai, Chi-Ting Liu, Yu Wang, Fu-Chieh Chang, Pei-Yuan Wu
Abstract: This study uses CAPM (Convex Adversarial Polytope for Maxpool-based CNN) to improve the verified bound for general purpose maxpool-based convolutional neural networks (CNNs) under bounded norm adversarial perturbations. The maxpool function is decomposed as a series of ReLU functions to extend the convex relaxation technique to maxpool functions, by which the verified bound can be efficiently computed through a dual network. The experimental results demonstrate that this technique allows the state-of-the-art verification precision for maxpool-based CNNs and involves a much lower computational cost than current verification methods, such as DeepZ, DeepPoly and PRIMA. This method is also applicable to large-scale CNNs, which previous studies show to be often computationally prohibitively expensive. Under certain circumstances, CAPM is 40-times, 20-times or twice as fast and give a significantly higher verification bound (CAPM 98% vs. PRIMA 76%/DeepPoly 73%/DeepZ 8%) as compared to PRIMA/DeepPoly/DeepZ. Furthermore, we additionally present the time complexity of our algorithm as $O(W^2NK)$, where $W$ is the maximum width of the neural network, $N$ is the number of neurons, and $K$ is the size of the maxpool layer's kernel.
Authors: Dongwei Ren, Xinya Shu, Yu Li, Xiaohe Wu, Jin Li, Wangmeng Zuo
Abstract: For single image defocus deblurring, acquiring well-aligned training pairs (or training triplets), i.e., a defocus blurry image, an all-in-focus sharp image (and a defocus blur map), is a challenging task for developing effective deblurring models. Existing image defocus deblurring methods typically rely on training data collected by specialized imaging equipment, with the assumption that these pairs or triplets are perfectly aligned. However, in practical scenarios involving the collection of real-world data, direct acquisition of training triplets is infeasible, and training pairs inevitably encounter spatial misalignment issues. In this work, we introduce a reblurring-guided learning framework for single image defocus deblurring, enabling the learning of a deblurring network even with misaligned training pairs. By reconstructing spatially variant isotropic blur kernels, our reblurring module ensures spatial consistency between the deblurred image, the reblurred image and the input blurry image, thereby addressing the misalignment issue while effectively extracting sharp textures from the all-in-focus sharp image. Moreover, spatially variant blur can be derived from the reblurring module, and serve as pseudo supervision for defocus blur map during training, interestingly transforming training pairs into training triplets. To leverage this pseudo supervision, we propose a lightweight defocus blur estimator coupled with a fusion block, which enhances deblurring performance through seamless integration with state-of-the-art deblurring networks. Additionally, we have collected a new dataset for single image defocus deblurring (SDD) with typical misalignments, which not only validates our proposed method but also serves as a benchmark for future research.
Authors: Shih-Han Chou, Shivam Chandhok, James J. Little, Leonid Sigal
Abstract: With the advent of LLMs and variants, a flurry of research has emerged, analyzing the performance of such models across an array of tasks. While most studies focus on evaluating the capabilities of state-of-the-art (SoTA) Vision Language Models (VLMs) through task accuracy (e.g., visual question answering, grounding), our work explores the related but complementary aspect of consistency - the ability of a VLM to produce semantically similar or identical responses to semantically similar queries. We note that consistency is a fundamental prerequisite (necessary but not sufficient condition) for robustness and trust in VLMs. Armed with this perspective, we propose the MM-R3 benchmark, which allows us to analyze performance, in terms of consistency and accuracy, of SoTA VLMs on three tasks: Question Rephrasing, Image Restyling, and Context Reasoning. Our analysis reveals that consistency does not always align with accuracy, indicating that models with higher accuracy are not necessarily more consistent, and vice versa. Furthermore, we propose a simple yet effective mitigation strategy in the form of an adapter module trained to minimize inconsistency across prompts. With our proposed strategy, we are able to achieve absolute improvements of 5.7% and 12.5%, on average on widely used VLMs such as BLIP-2 and LLaVa 1.5M in terms of consistency over their existing counterparts.
Authors: Ao-Xiang Zhang, Yuan-Gen Wang, Yu Ran, Weixuan Tang, Qingxiao Guan, Chunsheng Yang
Abstract: The exponential surge in video traffic has intensified the imperative for Video Quality Assessment (VQA). Leveraging cutting-edge architectures, current VQA models have achieved human-comparable accuracy. However, recent studies have revealed the vulnerability of existing VQA models against adversarial attacks. To establish a reliable and practical assessment system, a secure VQA model capable of resisting such malicious attacks is urgently demanded. Unfortunately, no attempt has been made to explore this issue. This paper first attempts to investigate general adversarial defense principles, aiming at endowing existing VQA models with security. Specifically, we first introduce random spatial grid sampling on the video frame for intra-frame defense. Then, we design pixel-wise randomization through a guardian map, globally neutralizing adversarial perturbations. Meanwhile, we extract temporal information from the video sequence as compensation for inter-frame defense. Building upon these principles, we present a novel VQA framework from the security-oriented perspective, termed SecureVQA. Extensive experiments indicate that SecureVQA sets a new benchmark in security while achieving competitive VQA performance compared with state-of-the-art models. Ablation studies delve deeper into analyzing the principles of SecureVQA, demonstrating their generalization and contributions to the security of leading VQA models.
Authors: Yuanhao Cai, He Zhang, Kai Zhang, Yixun Liang, Mengwei Ren, Fujun Luan, Qing Liu, Soo Ye Kim, Jianming Zhang, Zhifei Zhang, Yuqian Zhou, Yulun Zhang, Xiaokang Yang, Zhe Lin, Alan Yuille
Abstract: Existing feedforward image-to-3D methods mainly rely on 2D multi-view diffusion models that cannot guarantee 3D consistency. These methods easily collapse when changing the prompt view direction and mainly handle object-centric cases. In this paper, we propose a novel single-stage 3D diffusion model, DiffusionGS, for object generation and scene reconstruction from a single view. DiffusionGS directly outputs 3D Gaussian point clouds at each timestep to enforce view consistency and allow the model to generate robustly given prompt views of any directions, beyond object-centric inputs. Plus, to improve the capability and generality of DiffusionGS, we scale up 3D training data by developing a scene-object mixed training strategy. Experiments show that DiffusionGS yields improvements of 2.20 dB/23.25 and 1.34 dB/19.16 in PSNR/FID for objects and scenes than the state-of-the-art methods, without depth estimator. Plus, our method enjoys over 5$\times$ faster speed ($\sim$6s on an A100 GPU). Our Project page at https://caiyuanhao1998.github.io/project/DiffusionGS/ shows the video and interactive results.
Authors: Qihan Huang, Weilong Dai, Jinlong Liu, Wanggui He, Hao Jiang, Mingli Song, Jie Song
Abstract: Finetuning-free personalized image generation can synthesize customized images without test-time finetuning, attracting wide research interest owing to its high efficiency. Current finetuning-free methods simply adopt a single training stage with a simple image reconstruction task, and they typically generate low-quality images inconsistent with the reference images during test-time. To mitigate this problem, inspired by the recent DPO (i.e., direct preference optimization) technique, this work proposes an additional training stage to improve the pre-trained personalized generation models. However, traditional DPO only determines the overall superiority or inferiority of two samples, which is not suitable for personalized image generation because the generated images are commonly inconsistent with the reference images only in some local image patches. To tackle this problem, this work proposes PatchDPO that estimates the quality of image patches within each generated image and accordingly trains the model. To this end, PatchDPO first leverages the pre-trained vision model with a proposed self-supervised training method to estimate the patch quality. Next, PatchDPO adopts a weighted training approach to train the model with the estimated patch quality, which rewards the image patches with high quality while penalizing the image patches with low quality. Experiment results demonstrate that PatchDPO significantly improves the performance of multiple pre-trained personalized generation models, and achieves state-of-the-art performance on both single-object and multi-object personalized image generation. Our code is available at https://github.com/hqhQAQ/PatchDPO.
Authors: Zijian Zhao, Zhijie Cai, Tingwei Chen, Xiaoyang Li, Hang Li, Qimei Chen, Guangxu Zhu
Abstract: Wireless sensing has recently found widespread applications in diverse environments, including homes, offices, and public spaces. By analyzing patterns in channel state information (CSI), it is possible to infer human actions for tasks such as person identification, gesture recognition, and fall detection. However, CSI is highly sensitive to environmental changes, where even minor alterations can significantly distort the CSI patterns. This sensitivity often leads to performance degradation or outright failure when applying wireless sensing models trained in one environment to another. To address this challenge, Domain Alignment (DAL) has been widely adopted for cross-domain classification tasks, as it focuses on aligning the global distributions of the source and target domains in feature space. Despite its popularity, DAL often neglects inter-category relationships, which can lead to misalignment between categories across domains, even when global alignment is achieved. To overcome these limitations, we propose K-Nearest Neighbors Maximum Mean Discrepancy (KNN-MMD), a novel few-shot method for cross-domain wireless sensing. Our approach begins by constructing a help set using KNN from the target domain, enabling local alignment between the source and target domains within each category using MMD. Additionally, we address a key instability issue commonly observed in cross-domain methods, where model performance fluctuates sharply between epochs. Further, most existing methods struggle to determine an optimal stopping point during training due to the absence of labeled data from the target domain. Our method resolves this by excluding the support set from the target domain during training and employing it as a validation set to determine the stopping criterion.The dataset and code are publicly available at https://github.com/RS2002/KNN-MMD .
Authors: Connor Malone, Somayeh Hussaini, Tobias Fischer, Michael Milford
Abstract: Visual Place Recognition (VPR) enables coarse localization by comparing query images to a reference database of geo-tagged images. Recent breakthroughs in deep learning architectures and training regimes have led to methods with improved robustness to factors like environment appearance change, but with the downside that the required training and/or matching compute scales with the number of distinct environmental conditions encountered. Here, we propose Hyperdimensional One Place Signatures (HOPS) to simultaneously improve the performance, compute and scalability of these state-of-the-art approaches by fusing the descriptors from multiple reference sets captured under different conditions. HOPS scales to any number of environmental conditions by leveraging the Hyperdimensional Computing framework. Extensive evaluations demonstrate that our approach is highly generalizable and consistently improves recall performance across all evaluated VPR methods and datasets by large margins. Arbitrarily fusing reference images without compute penalty enables numerous other useful possibilities, three of which we demonstrate here: descriptor dimensionality reduction with no performance penalty, stacking synthetic images, and coarse localization to an entire traverse or environmental section.
Authors: Aleksandar Cvejic, Abdelrahman Eldesokey, Peter Wonka
Abstract: We present the first text-based image editing approach for object parts based on pre-trained diffusion models. Diffusion-based image editing approaches capitalized on the deep understanding of diffusion models of image semantics to perform a variety of edits. However, existing diffusion models lack sufficient understanding of many object parts, hindering fine-grained edits requested by users. To address this, we propose to expand the knowledge of pre-trained diffusion models to allow them to understand various object parts, enabling them to perform fine-grained edits. We achieve this by learning special textual tokens that correspond to different object parts through an efficient token optimization process. These tokens are optimized to produce reliable localization masks at each inference step to localize the editing region. Leveraging these masks, we design feature-blending and adaptive thresholding strategies to execute the edits seamlessly. To evaluate our approach, we establish a benchmark and an evaluation protocol for part editing. Experiments show that our approach outperforms existing editing methods on all metrics and is preferred by users 66-90% of the time in conducted user studies.
Authors: Hongyu An, Xinfeng Zhang, Shijie Zhao, Li Zhang, Ruiqin Xiong
Abstract: Due to storage and bandwidth limitations, videos transmitted over the Internet often exhibit low quality, characterized by low-resolution and compression artifacts. Although video super-resolution (VSR) is an efficient video enhancing technique, existing VSR methods focus less on compressed videos. Consequently, directly applying general VSR approaches fails to improve practical videos with compression artifacts, especially when frames are highly compressed at a low bit rate. The inevitable quantization information loss complicates the reconstruction of texture details. Recently, diffusion models have shown superior performance in low-level visual tasks. Leveraging the high-realism generation capability of diffusion models, we propose a novel method that exploits the priors of pre-trained diffusion models for compressed VSR. To mitigate spatial distortions and refine temporal consistency, we introduce a Spatial Degradation-Aware and Temporal Consistent (SDATC) diffusion model. Specifically, we incorporate a distortion control module (DCM) to modulate diffusion model inputs, thereby minimizing the impact of noise from low-quality frames on the generation stage. Subsequently, the diffusion model performs a denoising process to generate details, guided by a fine-tuned compression-aware prompt module (CAPM) and a spatio-temporal attention module (STAM). CAPM dynamically encodes compression-related information into prompts, enabling the sampling process to adapt to different degradation levels. Meanwhile, STAM extends the spatial attention mechanism into the spatio-temporal dimension, effectively capturing temporal correlations. Additionally, we utilize optical flow-based alignment during each denoising step to enhance the smoothness of output videos. Extensive experimental results on benchmark datasets demonstrate the effectiveness of our proposed modules in restoring compressed videos.
Authors: Liying Yang, Chen Liu, Zhenwei Zhu, Ajian Liu, Hui Ma, Jian Nong, Yanyan Liang
Abstract: Recently, the generation of dynamic 3D objects from a video has shown impressive results. Existing methods directly optimize Gaussians using whole information in frames. However, when dynamic regions are interwoven with static regions within frames, particularly if the static regions account for a large proportion, existing methods often overlook information in dynamic regions and are prone to overfitting on static regions. This leads to producing results with blurry textures. We consider that decoupling dynamic-static features to enhance dynamic representations can alleviate this issue. Thus, we propose a dynamic-static feature decoupling module (DSFD). Along temporal axes, it regards the regions of current frame features that possess significant differences relative to reference frame features as dynamic features. Conversely, the remaining parts are the static features. Then, we acquire decoupled features driven by dynamic features and current frame features. Moreover, to further enhance the dynamic representation of decoupled features from different viewpoints and ensure accurate motion prediction, we design a temporal-spatial similarity fusion module (TSSF). Along spatial axes, it adaptively selects similar information of dynamic regions. Hinging on the above, we construct a novel approach, DS4D. Experimental results verify our method achieves state-of-the-art (SOTA) results in video-to-4D. In addition, the experiments on a real-world scenario dataset demonstrate its effectiveness on the 4D scene. Our code will be publicly available.
Authors: Ahmed Heakl, Abdullah Sohail, Mukul Ranjan, Rania Hossam, Ghazi Shazan Ahmad, Mohamed El-Geish, Omar Maher, Zhiqiang Shen, Fahad Khan, Salman Khan
Abstract: With the growing adoption of Retrieval-Augmented Generation (RAG) in document processing, robust text recognition has become increasingly critical for knowledge extraction. While OCR (Optical Character Recognition) for English and other languages benefits from large datasets and well-established benchmarks, Arabic OCR faces unique challenges due to its cursive script, right-to-left text flow, and complex typographic and calligraphic features. We present KITAB-Bench, a comprehensive Arabic OCR benchmark that fills the gaps in current evaluation systems. Our benchmark comprises 8,809 samples across 9 major domains and 36 sub-domains, encompassing diverse document types including handwritten text, structured tables, and specialized coverage of 21 chart types for business intelligence. Our findings show that modern vision-language models (such as GPT-4o, Gemini, and Qwen) outperform traditional OCR approaches (like EasyOCR, PaddleOCR, and Surya) by an average of 60% in Character Error Rate (CER). Furthermore, we highlight significant limitations of current Arabic OCR models, particularly in PDF-to-Markdown conversion, where the best model Gemini-2.0-Flash achieves only 65% accuracy. This underscores the challenges in accurately recognizing Arabic text, including issues with complex fonts, numeral recognition errors, word elongation, and table structure detection. This work establishes a rigorous evaluation framework that can drive improvements in Arabic document analysis methods and bridge the performance gap with English OCR technologies.
Authors: Yitao Zhu, Yuan Yin, Jiaming Li, Mengjie Xu, Zihao Zhao, Honglin Xiong, Sheng Wang, Qian Wang
Abstract: The adoption of visual foundation models has become a common practice in computer-aided diagnosis (CAD). While these foundation models provide a viable solution for creating generalist medical AI, privacy concerns make it difficult to pre-train or continuously update such models across multiple domains and datasets, leading many studies to focus on specialist models. To address this challenge, we propose Med-LEGO, a training-free framework that enables the seamless integration or updating of a generalist CAD model by combining multiple specialist models, similar to assembling LEGO bricks. Med-LEGO enhances LoRA (low-rank adaptation) by incorporating singular value decomposition (SVD) to efficiently capture the domain expertise of each specialist model with minimal additional parameters. By combining these adapted weights through simple operations, Med-LEGO allows for the easy integration or modification of specific diagnostic capabilities without the need for original data or retraining. Finally, the combined model can be further adapted to new diagnostic tasks, making it a versatile generalist model. Our extensive experiments demonstrate that Med-LEGO outperforms existing methods in both cross-domain and in-domain medical tasks while using only 0.18% of full model parameters. These merged models show better convergence and generalization to new tasks, providing an effective path toward generalist medical AI.
Authors: Xuanhan Wang, Huimin Deng, Lianli Gao, Jingkuan Song
Abstract: Human-centric visual perception (HVP) has recently achieved remarkable progress due to advancements in large-scale self-supervised pretraining (SSP). However, existing HVP models face limitations in adapting to real-world applications, which require general visual patterns for downstream tasks while maintaining computationally sustainable costs to ensure compatibility with edge devices. These limitations primarily arise from two issues: 1) the pretraining objectives focus solely on specific visual patterns, limiting the generalizability of the learned patterns for diverse downstream tasks; and 2) HVP models often exhibit excessively large model sizes, making them incompatible with real-world applications.To address these limitations, we introduce Scale-Aware Image Pretraining (SAIP), a novel SSP framework pretraining lightweight vision models to acquire general patterns for HVP. Specifically, SAIP incorporates three learning objectives based on the principle of cross-scale consistency: 1) Cross-scale Matching (CSM) which contrastively learns image-level invariant patterns from multi-scale single-person images; 2) Cross-scale Reconstruction (CSR) which learns pixel-level consistent visual structures from multi-scale masked single-person images; and 3) Cross-scale Search (CSS) which learns to capture diverse patterns from multi-scale multi-person images. Three objectives complement one another, enabling lightweight models to learn multi-scale generalizable patterns essential for HVP downstream tasks.Extensive experiments conducted across 12 HVP datasets demonstrate that SAIP exhibits remarkable generalization capabilities across 9 human-centric vision tasks. Moreover, it achieves significant performance improvements over existing methods, with gains of 3%-13% in single-person discrimination tasks, 1%-11% in dense prediction tasks, and 1%-6% in multi-person visual understanding tasks.
Authors: Ruichen Chen, Keith G. Mills, Di Niu
Abstract: Diffusion Models (DM) have revolutionized the text-to-image visual generation process. However, the large computational cost and model footprint of DMs hinders practical deployment, especially on edge devices. Post-training quantization (PTQ) is a lightweight method to alleviate these burdens without the need for training or fine-tuning. While recent DM PTQ methods achieve W4A8 on integer-based PTQ, two key limitations remain: First, while most existing DM PTQ methods evaluate on classical DMs like Stable Diffusion XL, 1.5 or earlier, which use convolutional U-Nets, newer Diffusion Transformer (DiT) models like the PixArt series, Hunyuan and others adopt fundamentally different transformer backbones to achieve superior image synthesis. Second, integer (INT) quantization is prevailing in DM PTQ but doesn't align well with the network weight and activation distribution, while Floating-Point Quantization (FPQ) is still under-investigated, yet it holds the potential to better align the weight and activation distributions in low-bit settings for DiT. In response, we introduce FP4DiT, a PTQ method that leverages FPQ to achieve W4A6 quantization. Specifically, we extend and generalize the Adaptive Rounding PTQ technique to adequately calibrate weight quantization for FPQ and demonstrate that DiT activations depend on input patch data, necessitating robust online activation quantization techniques. Experimental results demonstrate that FP4DiT outperforms integer-based PTQ at W4A6 and W4A8 precision and generates convincing visual content on PixArt-$\alpha$, PixArt-$\Sigma$ and Hunyuan in terms of several T2I metrics such as HPSv2 and CLIP.
Authors: Aniek Eijpe, Soufyan Lakbir, Melis Erdal Cesur, Sara P. Oliveira, Sanne Abeln, Wilson Silva
Abstract: To improve the prediction of cancer survival using whole-slide images and transcriptomics data, it is crucial to capture both modality-shared and modality-specific information. However, multimodal frameworks often entangle these representations, limiting interpretability and potentially suppressing discriminative features. To address this, we propose Disentangled and Interpretable Multimodal Attention Fusion (DIMAF), a multimodal framework that separates the intra- and inter-modal interactions within an attention-based fusion mechanism to learn distinct modality-specific and modality-shared representations. We introduce a loss based on Distance Correlation to promote disentanglement between these representations and integrate Shapley additive explanations to assess their relative contributions to survival prediction. We evaluate DIMAF on four public cancer survival datasets, achieving a relative average improvement of 1.85% in performance and 23.7% in disentanglement compared to current state-of-the-art multimodal models. Beyond improved performance, our interpretable framework enables a deeper exploration of the underlying interactions between and within modalities in cancer biology.
Authors: Zeng-Hui Zhu, Wei Lu, Si-Bao Chen, Chris H. Q. Ding, Jin Tang, Bin Luo
Abstract: Remote Sensing Image Dehazing (RSID) poses significant challenges in real-world scenarios due to the complex atmospheric conditions and severe color distortions that degrade image quality. The scarcity of real-world remote sensing hazy image pairs has compelled existing methods to rely primarily on synthetic datasets. However, these methods struggle with real-world applications due to the inherent domain gap between synthetic and real data. To address this, we introduce Real-World Remote Sensing Hazy Image Dataset (RRSHID), the first large-scale dataset featuring real-world hazy and dehazed image pairs across diverse atmospheric conditions. Based on this, we propose MCAF-Net, a novel framework tailored for real-world RSID. Its effectiveness arises from three innovative components: Multi-branch Feature Integration Block Aggregator (MFIBA), which enables robust feature extraction through cascaded integration blocks and parallel multi-branch processing; Color-Calibrated Self-Supervised Attention Module (CSAM), which mitigates complex color distortions via self-supervised learning and attention-guided refinement; and Multi-Scale Feature Adaptive Fusion Module (MFAFM), which integrates features effectively while preserving local details and global context. Extensive experiments validate that MCAF-Net demonstrates state-of-the-art performance in real-world RSID, while maintaining competitive performance on synthetic datasets. The introduction of RRSHID and MCAF-Net sets new benchmarks for real-world RSID research, advancing practical solutions for this complex task. The code and dataset are publicly available at https://github.com/lwCVer/RRSHID.
Authors: Zizhi Chen, Minghao Han, Xukun Zhang, Shuwei Ma, Tao Liu, Xing Wei, Lihua Zhang
Abstract: Multimodal learning combining pathology images and genomic sequences enhances cancer survival analysis but faces clinical implementation barriers due to limited access to genomic sequencing in under-resourced regions. To enable survival prediction using only whole-slide images (WSI), we propose the Visual-Genomic Answering-Guided Transformer (VGAT), a framework integrating Visual Question Answering (VQA) techniques for genomic modality reconstruction. By adapting VQA's text feature extraction approach, we derive stable genomic representations that circumvent dimensionality challenges in raw genomic data. Simultaneously, a cluster-based visual prompt module selectively enhances discriminative WSI patches, addressing noise from unfiltered image regions. Evaluated across five TCGA datasets, VGAT outperforms existing WSI-only methods, demonstrating the viability of genomic-informed inference without sequencing. This approach bridges multimodal research and clinical feasibility in resource-constrained settings. The code link is https://github.com/CZZZZZZZZZZZZZZZZZ/VGAT.
Authors: Joao Pereira, Vasco Lopes, David Semedo, Joao Neves
Abstract: Large Vision-Language Models (LVLMs) demonstrate remarkable performance in short-video tasks such as video question answering, but struggle in long-video understanding. The linear frame sampling strategy, conventionally used by LVLMs, fails to account for the non-linear distribution of key events in video data, often introducing redundant or irrelevant information in longer contexts while risking the omission of critical events in shorter ones. To address this, we propose SelfReS, a non-linear spatiotemporal self-reflective sampling method that dynamically selects key video fragments based on user prompts. Unlike prior approaches, SelfReS leverages the inherently sparse attention maps of LVLMs to define reflection tokens, enabling relevance-aware token selection without requiring additional training or external modules. Experiments demonstrate that SelfReS can be seamlessly integrated into strong base LVLMs, improving long-video task accuracy and achieving up to 46% faster inference speed within the same GPU memory budget.
Authors: Linfeng Tang, Yeda Wang, Meiqi Gong, Zizhuo Li, Yuxin Deng, Xunpeng Yi, Chunyu Li, Han Xu, Hao Zhang, Jiayi Ma
Abstract: Compared to images, videos better align with real-world acquisition scenarios and possess valuable temporal cues. However, existing multi-sensor fusion research predominantly integrates complementary context from multiple images rather than videos. This primarily stems from two factors: 1) the scarcity of large-scale multi-sensor video datasets, limiting research in video fusion, and 2) the inherent difficulty of jointly modeling spatial and temporal dependencies in a unified framework. This paper proactively compensates for the dilemmas. First, we construct M3SVD, a benchmark dataset with $220$ temporally synchronized and spatially registered infrared-visible video pairs comprising 153,797 frames, filling the data gap for the video fusion community. Secondly, we propose VideoFusion, a multi-modal video fusion model that fully exploits cross-modal complementarity and temporal dynamics to generate spatio-temporally coherent videos from (potentially degraded) multi-modal inputs. Specifically, 1) a differential reinforcement module is developed for cross-modal information interaction and enhancement, 2) a complete modality-guided fusion strategy is employed to adaptively integrate multi-modal features, and 3) a bi-temporal co-attention mechanism is devised to dynamically aggregate forward-backward temporal contexts to reinforce cross-frame feature representations. Extensive experiments reveal that VideoFusion outperforms existing image-oriented fusion paradigms in sequential scenarios, effectively mitigating temporal inconsistency and interference.
Authors: Qihan Huang, Weilong Dai, Jinlong Liu, Wanggui He, Hao Jiang, Mingli Song, Jingyuan Chen, Chang Yao, Jie Song
Abstract: MLLM reasoning has drawn widespread research for its excellent problem-solving capability. Current reasoning methods fall into two types: PRM, which supervises the intermediate reasoning steps, and ORM, which supervises the final results. Recently, DeepSeek-R1 has challenged the traditional view that PRM outperforms ORM, which demonstrates strong generalization performance using an ORM method (i.e., GRPO). However, current MLLM's GRPO algorithms still struggle to handle challenging and complex multimodal reasoning tasks (e.g., mathematical reasoning). In this work, we reveal two problems that impede the performance of GRPO on the MLLM: Low data utilization and Text-bias. Low data utilization refers to that GRPO cannot acquire positive rewards to update the MLLM on difficult samples, and text-bias is a phenomenon that the MLLM bypasses image condition and solely relies on text condition for generation after GRPO training. To tackle these problems, this work proposes Hint-GRPO that improves data utilization by adaptively providing hints for samples of varying difficulty, and text-bias calibration that mitigates text-bias by calibrating the token prediction logits with image condition in test-time. Experiment results on three base MLLMs across eleven datasets demonstrate that our proposed methods advance the reasoning capability of original MLLM by a large margin, exhibiting superior performance to existing MLLM reasoning methods. Our code is available at https://github.com/hqhQAQ/Hint-GRPO.
Authors: Yechan Kim, DongHo Yoon, SooYeon Kim, Moongu Jeon
Abstract: Labeling errors in remote sensing (RS) image segmentation datasets often remain implicit and subtle due to ambiguous class boundaries, mixed pixels, shadows, complex terrain features, and subjective annotator bias. Furthermore, the scarcity of annotated RS data due to high image acquisition and labeling costs complicates training noise-robust models. While sophisticated mechanisms such as label selection or noise correction might address this issue, they tend to increase training time and add implementation complexity. In this letter, we propose NSegment-a simple yet effective data augmentation solution to mitigate this issue. Unlike traditional methods, it applies elastic transformations only to segmentation labels, varying deformation intensity per sample in each training epoch to address annotation inconsistencies. Experimental results demonstrate that our approach improves the performance of RS image segmentation on various state-of-the-art models.
Authors: Xinjie Zhang, Jintao Guo, Shanshan Zhao, Minghao Fu, Lunhao Duan, Jiakui Hu, Yong Xien Chng, Guo-Hua Wang, Qing-Guo Chen, Zhao Xu, Weihua Luo, Kaifu Zhang
Abstract: Recent years have seen remarkable progress in both multimodal understanding models and image generation models. Despite their respective successes, these two domains have evolved independently, leading to distinct architectural paradigms: While autoregressive-based architectures have dominated multimodal understanding, diffusion-based models have become the cornerstone of image generation. Recently, there has been growing interest in developing unified frameworks that integrate these tasks. The emergence of GPT-4o's new capabilities exemplifies this trend, highlighting the potential for unification. However, the architectural differences between the two domains pose significant challenges. To provide a clear overview of current efforts toward unification, we present a comprehensive survey aimed at guiding future research. First, we introduce the foundational concepts and recent advancements in multimodal understanding and text-to-image generation models. Next, we review existing unified models, categorizing them into three main architectural paradigms: diffusion-based, autoregressive-based, and hybrid approaches that fuse autoregressive and diffusion mechanisms. For each category, we analyze the structural designs and innovations introduced by related works. Additionally, we compile datasets and benchmarks tailored for unified models, offering resources for future exploration. Finally, we discuss the key challenges facing this nascent field, including tokenization strategy, cross-modal attention, and data. As this area is still in its early stages, we anticipate rapid advancements and will regularly update this survey. Our goal is to inspire further research and provide a valuable reference for the community. The references associated with this survey are available on GitHub (https://github.com/AIDC-AI/Awesome-Unified-Multimodal-Models).
URLs: https://github.com/AIDC-AI/Awesome-Unified-Multimodal-Models).
Authors: Jialei Chen, Xu Zheng, Dongyue Li, Chong Yi, Seigo Ito, Danda Pani Paudel, Luc Van Gool, Hiroshi Murase, Daisuke Deguchi
Abstract: Zero-shot Semantic Segmentation (ZSS) aims to segment categories that are not annotated during training. While fine-tuning vision-language models has achieved promising results, these models often overfit to seen categories due to the lack of supervision for unseen classes. As an alternative to fully supervised approaches, query-based segmentation has shown great latent in ZSS, as it enables object localization without relying on explicit labels. However, conventional Hungarian matching, a core component in query-based frameworks, needs full supervision and often misclassifies unseen categories as background in the setting of ZSS. To address this issue, we propose Split Matching (SM), a novel assignment strategy that decouples Hungarian matching into two components: one for seen classes in annotated regions and another for latent classes in unannotated regions (referred to as unseen candidates). Specifically, we partition the queries into seen and candidate groups, enabling each to be optimized independently according to its available supervision. To discover unseen candidates, we cluster CLIP dense features to generate pseudo masks and extract region-level embeddings using CLS tokens. Matching is then conducted separately for the two groups based on both class-level similarity and mask-level consistency. Additionally, we introduce a Multi-scale Feature Enhancement (MFE) module that refines decoder features through residual multi-scale aggregation, improving the model's ability to capture spatial details across resolutions. SM is the first to introduce decoupled Hungarian matching under the inductive ZSS setting, and achieves state-of-the-art performance on two standard benchmarks.
Authors: Daoxi Cao, Hangbei Cheng, Yijin Li, Ruolin Zhou, Xuehan Zhang, Xinyi Li, Binwei Li, Xuancheng Gu, Jianan Zhang, Xueyu Liu, Yongfei Wu
Abstract: Whole-slide images (WSIs) are critical for cancer diagnosis due to their ultra-high resolution and rich semantic content. However, their massive size and the limited availability of fine-grained annotations pose substantial challenges for conventional supervised learning. We propose DSAGL (Dual-Stream Attention-Guided Learning), a novel weakly supervised classification framework that combines a teacher-student architecture with a dual-stream design. DSAGL explicitly addresses instance-level ambiguity and bag-level semantic consistency by generating multi-scale attention-based pseudo labels and guiding instance-level learning. A shared lightweight encoder (VSSMamba) enables efficient long-range dependency modeling, while a fusion-attentive module (FASA) enhances focus on sparse but diagnostically relevant regions. We further introduce a hybrid loss to enforce mutual consistency between the two streams. Experiments on CIFAR-10, NCT-CRC, and TCGA-Lung datasets demonstrate that DSAGL consistently outperforms state-of-the-art MIL baselines, achieving superior discriminative performance and robustness under weak supervision.
Authors: Nokimul Hasan Arif, Shadman Rabby, Md Hefzul Hossain Papon, Sabbir Ahmed
Abstract: Visual hallucinations in Large Language Models (LLMs), where the model generates responses that are inconsistent with the visual input, pose a significant challenge to their reliability, particularly in contexts where precise and trustworthy outputs are critical. Current research largely emphasizes post-hoc correction or model-specific fine-tuning strategies, with limited exploration of preprocessing techniques to address hallucination issues at the input stage. This study presents a novel ensemble-based preprocessing framework that adaptively selects the most appropriate filtering approach -- noise reduced (NR), edge enhanced (EE), or unaltered input (org) based on the type of question posed, resulting into reduced hallucination without requiring any modifications to the underlying model architecture or training pipeline. Evaluated on the `HaloQuest' dataset -- a benchmark designed to test multimodal reasoning on visually complex inputs, our method achieves a 44.3% reduction in hallucination rates, as measured by Natural Language Inference (NLI) scores using SelfCheckGPT. This demonstrates that intelligent input conditioning alone can significantly enhance factual grounding in LLM responses. The findings highlight the importance of adaptive preprocessing techniques in mitigating hallucinations, paving the way for more reliable multimodal systems capable of addressing real-world challenges.
Authors: Nick Jiang, Amil Dravid, Alexei Efros, Yossi Gandelsman
Abstract: We investigate the mechanism underlying a previously identified phenomenon in Vision Transformers -- the emergence of high-norm tokens that lead to noisy attention maps. We observe that in multiple models (e.g., CLIP, DINOv2), a sparse set of neurons is responsible for concentrating high-norm activations on outlier tokens, leading to irregular attention patterns and degrading downstream visual processing. While the existing solution for removing these outliers involves retraining models from scratch with additional learned register tokens, we use our findings to create a training-free approach to mitigate these artifacts. By shifting the high-norm activations from our discovered register neurons into an additional untrained token, we can mimic the effect of register tokens on a model already trained without registers. We demonstrate that our method produces cleaner attention and feature maps, enhances performance over base models across multiple downstream visual tasks, and achieves results comparable to models explicitly trained with register tokens. We then extend test-time registers to off-the-shelf vision-language models to improve their interpretability. Our results suggest that test-time registers effectively take on the role of register tokens at test-time, offering a training-free solution for any pre-trained model released without them.
Authors: Zhijing Wan, Zhixiang Wang, Zheng Wang, Xin Xu, Shin'ichi Satoh
Abstract: One-shot subset selection serves as an effective tool to reduce deep learning training costs by identifying an informative data subset based on the information extracted by an information extractor (IE). Traditional IEs, typically pre-trained on the target dataset, are inherently dataset-dependent. Foundation models (FMs) offer a promising alternative, potentially mitigating this limitation. This work investigates two key questions: (1) Can FM-based subset selection outperform traditional IE-based methods across diverse datasets? (2) Do all FMs perform equally well as IEs for subset selection? Extensive experiments uncovered surprising insights: FMs consistently outperform traditional IEs on fine-grained datasets, whereas their advantage diminishes on coarse-grained datasets with noisy labels. Motivated by these finding, we propose RAM-APL (RAnking Mean-Accuracy of Pseudo-class Labels), a method tailored for fine-grained image datasets. RAM-APL leverages multiple FMs to enhance subset selection by exploiting their complementary strengths. Our approach achieves state-of-the-art performance on fine-grained datasets, including Oxford-IIIT Pet, Food-101, and Caltech-UCSD Birds-200-2011.
Authors: Fei Zhou
Abstract: Remote sensing change detection is used in urban planning, terrain analysis, and environmental monitoring by analyzing feature changes in the same area over time. In this paper, we propose a large language model (LLM) augmented inference approach (SegChange-R1), which enhances the detection capability by integrating textual descriptive information and guides the model to focus on relevant change regions, accelerating convergence. We designed a linear attention-based spatial transformation module (BEV) to address modal misalignment by unifying features from different times into a BEV space. Furthermore, we introduce DVCD, a novel dataset for building change detection from UAV viewpoints. Experiments on four widely-used datasets demonstrate significant improvements over existing method The code and pre-trained models are available in {https://github.com/Yu-Zhouz/SegChange-R1}.
Authors: Jisheng Dang, Huilin Song, Junbin Xiao, Bimei Wang, Han Peng, Haoxuan Li, Xun Yang, Meng Wang, Tat-Seng Chua
Abstract: Grounded Video Question Answering (Grounded VideoQA) requires aligning textual answers with explicit visual evidence. However, modern multimodal models often rely on linguistic priors and spurious correlations, resulting in poorly grounded predictions. In this work, we propose MUPA, a cooperative MUlti-Path Agentic approach that unifies video grounding, question answering, answer reflection and aggregation to tackle Grounded VideoQA. MUPA features three distinct reasoning paths on the interplay of grounding and QA agents in different chronological orders, along with a dedicated reflection agent to judge and aggregate the multi-path results to accomplish consistent QA and grounding. This design markedly improves grounding fidelity without sacrificing answer accuracy. Despite using only 2B parameters, our method outperforms all 7B-scale competitors. When scaled to 7B parameters, MUPA establishes new state-of-the-art results, with Acc@GQA of 30.3% and 47.4% on NExT-GQA and DeVE-QA respectively, demonstrating MUPA' effectiveness towards trustworthy video-language understanding. Our code is available in https://github.com/longmalongma/MUPA.
Authors: Quoc-Duy Tran, Anh-Tuan Vo, Dinh-Khoi Vo, Tam V. Nguyen, Minh-Triet Tran, Trung-Nghia Le
Abstract: Humans possess a unique ability to perceive meaningful patterns in ambiguous stimuli, a cognitive phenomenon known as pareidolia. This paper introduces Shape2Animal framework to mimics this imaginative capacity by reinterpreting natural object silhouettes, such as clouds, stones, or flames, as plausible animal forms. Our automated framework first performs open-vocabulary segmentation to extract object silhouette and interprets semantically appropriate animal concepts using vision-language models. It then synthesizes an animal image that conforms to the input shape, leveraging text-to-image diffusion model and seamlessly blends it into the original scene to generate visually coherent and spatially consistent compositions. We evaluated Shape2Animal on a diverse set of real-world inputs, demonstrating its robustness and creative potential. Our Shape2Animal can offer new opportunities for visual storytelling, educational content, digital art, and interactive media design. Our project page is here: https://shape2image.github.io
Authors: Qin Ren, Yifan Wang, Ruogu Fang, Haibin Ling, Chenyu You
Abstract: Survival prediction using whole slide images (WSIs) can be formulated as a multiple instance learning (MIL) problem. However, existing MIL methods often fail to explicitly capture pathological heterogeneity within WSIs, both globally -- through long-tailed morphological distributions, and locally through -- tile-level prediction uncertainty. Optimal transport (OT) provides a principled way of modeling such heterogeneity by incorporating marginal distribution constraints. Building on this insight, we propose OTSurv, a novel MIL framework from an optimal transport perspective. Specifically, OTSurv formulates survival predictions as a heterogeneity-aware OT problem with two constraints: (1) global long-tail constraint that models prior morphological distributions to avert both mode collapse and excessive uniformity by regulating transport mass allocation, and (2) local uncertainty-aware constraint that prioritizes high-confidence patches while suppressing noise by progressively raising the total transport mass. We then recast the initial OT problem, augmented by these constraints, into an unbalanced OT formulation that can be solved with an efficient, hardware-friendly matrix scaling algorithm. Empirically, OTSurv sets new state-of-the-art results across six popular benchmarks, achieving an absolute 3.6% improvement in average C-index. In addition, OTSurv achieves statistical significance in log-rank tests and offers high interpretability, making it a powerful tool for survival prediction in digital pathology. Our codes are available at https://github.com/Y-Research-SBU/OTSurv.
Authors: Hao Zhang, Haolan Xu, Chun Feng, Varun Jampani, Narendra Ahuja
Abstract: Skinning and rigging are fundamental components in animation, articulated object reconstruction, motion transfer, and 4D generation. Existing approaches predominantly rely on Linear Blend Skinning (LBS), due to its simplicity and differentiability. However, LBS introduces artifacts such as volume loss and unnatural deformations, and it fails to model elastic materials like soft tissues, fur, and flexible appendages (e.g., elephant trunks, ears, and fatty tissues). In this work, we propose PhysRig: a differentiable physics-based skinning and rigging framework that overcomes these limitations by embedding the rigid skeleton into a volumetric representation (e.g., a tetrahedral mesh), which is simulated as a deformable soft-body structure driven by the animated skeleton. Our method leverages continuum mechanics and discretizes the object as particles embedded in an Eulerian background grid to ensure differentiability with respect to both material properties and skeletal motion. Additionally, we introduce material prototypes, significantly reducing the learning space while maintaining high expressiveness. To evaluate our framework, we construct a comprehensive synthetic dataset using meshes from Objaverse, The Amazing Animals Zoo, and MixaMo, covering diverse object categories and motion patterns. Our method consistently outperforms traditional LBS-based approaches, generating more realistic and physically plausible results. Furthermore, we demonstrate the applicability of our framework in the pose transfer task highlighting its versatility for articulated object modeling.
Authors: Lingling Cai, Kang Zhao, Hangjie Yuan, Xiang Wang, Yingya Zhang, Kejie Huang
Abstract: The advent of Video Diffusion Transformers (Video DiTs) marks a milestone in video generation. However, directly applying existing video editing methods to Video DiTs often incurs substantial computational overhead, due to resource-intensive attention modification or finetuning. To alleviate this problem, we present DFVEdit, an efficient zero-shot video editing method tailored for Video DiTs. DFVEdit eliminates the need for both attention modification and fine-tuning by directly operating on clean latents via flow transformation. To be more specific, we observe that editing and sampling can be unified under the continuous flow perspective. Building upon this foundation, we propose the Conditional Delta Flow Vector (CDFV) -- a theoretically unbiased estimation of DFV -- and integrate Implicit Cross Attention (ICA) guidance as well as Embedding Reinforcement (ER) to further enhance editing quality. DFVEdit excels in practical efficiency, offering at least 20x inference speed-up and 85% memory reduction on Video DiTs compared to attention-engineering-based editing methods. Extensive quantitative and qualitative experiments demonstrate that DFVEdit can be seamlessly applied to popular Video DiTs (e.g., CogVideoX and Wan2.1), attaining state-of-the-art performance on structural fidelity, spatial-temporal consistency, and editing quality.
Authors: Akio Hayakawa, Masato Ishii, Takashi Shibuya, Yuki Mitsufuji
Abstract: We propose a novel step-by-step video-to-audio generation method that sequentially produces individual audio tracks, each corresponding to a specific sound event in the video. Our approach mirrors traditional Foley workflows, aiming to capture all sound events induced by a given video comprehensively. Each generation step is formulated as a guided video-to-audio synthesis task, conditioned on a target text prompt and previously generated audio tracks. This design is inspired by the idea of concept negation from prior compositional generation frameworks. To enable this guided generation, we introduce a training framework that leverages pre-trained video-to-audio models and eliminates the need for specialized paired datasets, allowing training on more accessible data. Experimental results demonstrate that our method generates multiple semantically distinct audio tracks for a single input video, leading to higher-quality composite audio synthesis than existing baselines.
Authors: Bang Gong, Luchao Qi, Jiaye Wu, Zhicheng Fu, Chunbo Song, David W. Jacobs, John Nicholson, Roni Sengupta
Abstract: We introduce the Aging Multiverse, a framework for generating multiple plausible facial aging trajectories from a single image, each conditioned on external factors such as environment, health, and lifestyle. Unlike prior methods that model aging as a single deterministic path, our approach creates an aging tree that visualizes diverse futures. To enable this, we propose a training-free diffusion-based method that balances identity preservation, age accuracy, and condition control. Our key contributions include attention mixing to modulate editing strength and a Simulated Aging Regularization strategy to stabilize edits. Extensive experiments and user studies demonstrate state-of-the-art performance across identity preservation, aging realism, and conditional alignment, outperforming existing editing and age-progression models, which often fail to account for one or more of the editing criteria. By transforming aging into a multi-dimensional, controllable, and interpretable process, our approach opens up new creative and practical avenues in digital storytelling, health education, and personalized visualization.
Authors: Wenzhou Lyu, Jialing Lin, Wenqi Ren, Ruihao Xia, Feng Qian, Yang Tang
Abstract: Commercial RGB-D cameras often produce noisy, incomplete depth maps for non-Lambertian objects. Traditional depth completion methods struggle to generalize due to the limited diversity and scale of training data. Recent advances exploit visual priors from pre-trained text-to-image diffusion models to enhance generalization in dense prediction tasks. However, we find that biases arising from training-inference mismatches in the vanilla diffusion framework significantly impair depth completion performance. Additionally, the lack of distinct visual features in non-Lambertian regions further hinders precise prediction. To address these issues, we propose \textbf{DidSee}, a diffusion-based framework for depth completion on non-Lambertian objects. First, we integrate a rescaled noise scheduler enforcing a zero terminal signal-to-noise ratio to eliminate signal leakage bias. Second, we devise a noise-agnostic single-step training formulation to alleviate error accumulation caused by exposure bias and optimize the model with a task-specific loss. Finally, we incorporate a semantic enhancer that enables joint depth completion and semantic segmentation, distinguishing objects from backgrounds and yielding precise, fine-grained depth maps. DidSee achieves state-of-the-art performance on multiple benchmarks, demonstrates robust real-world generalization, and effectively improves downstream tasks such as category-level pose estimation and robotic grasping.
Authors: Xiwei Xuan, Ziquan Deng, Kwan-Liu Ma
Abstract: Training-free open-vocabulary semantic segmentation (OVS) aims to segment images given a set of arbitrary textual categories without costly model fine-tuning. Existing solutions often explore attention mechanisms of pre-trained models, such as CLIP, or generate synthetic data and design complex retrieval processes to perform OVS. However, their performance is limited by the capability of reliant models or the suboptimal quality of reference sets. In this work, we investigate the largely overlooked data quality problem for this challenging dense scene understanding task, and identify that a high-quality reference set can significantly benefit training-free OVS. With this observation, we introduce a data-quality-oriented framework, comprising a data pipeline to construct a reference set with well-paired segment-text embeddings and a simple similarity-based retrieval to unveil the essential effect of data. Remarkably, extensive evaluations on ten benchmark datasets demonstrate that our method outperforms all existing training-free OVS approaches, highlighting the importance of data-centric design for advancing OVS without training. Our code is available at https://github.com/xiweix/ReME .
Authors: Hongbo Liu, Jingwen He, Yi Jin, Dian Zheng, Yuhao Dong, Fan Zhang, Ziqi Huang, Yinan He, Yangguang Li, Weichao Chen, Yu Qiao, Wanli Ouyang, Shengjie Zhao, Ziwei Liu
Abstract: Cinematography, the fundamental visual language of film, is essential for conveying narrative, emotion, and aesthetic quality. While recent Vision-Language Models (VLMs) demonstrate strong general visual understanding, their proficiency in comprehending the nuanced cinematic grammar embedded within individual shots remains largely unexplored and lacks robust evaluation. This critical gap limits both fine-grained visual comprehension and the precision of AI-assisted video generation. To address this, we introduce ShotBench, a comprehensive benchmark specifically designed for cinematic language understanding. It features over 3.5k expert-annotated QA pairs from images and video clips, meticulously curated from over 200 acclaimed (predominantly Oscar-nominated) films and spanning eight key cinematography dimensions. Our evaluation of 24 leading VLMs on ShotBench reveals their substantial limitations: even the top-performing model achieves less than 60% average accuracy, particularly struggling with fine-grained visual cues and complex spatial reasoning. To catalyze advancement in this domain, we construct ShotQA, a large-scale multimodal dataset comprising approximately 70k cinematic QA pairs. Leveraging ShotQA, we develop ShotVL through supervised fine-tuning and Group Relative Policy Optimization. ShotVL significantly outperforms all existing open-source and proprietary models on ShotBench, establishing new state-of-the-art performance. We open-source our models, data, and code to foster rapid progress in this crucial area of AI-driven cinematic understanding and generation.
Authors: Riccardo Zaccone, Sai Praneeth Karimireddy, Carlo Masone, Marco Ciccone
Abstract: Federated Learning (FL) has emerged as the state-of-the-art approach for learning from decentralized data in privacy-constrained scenarios.However, system and statistical challenges hinder its real-world applicability, requiring efficient learning from edge devices and robustness to data heterogeneity. Despite significant research efforts, existing approaches often degrade severely due to the joint effect of heterogeneity and partial client participation. In particular, while momentum appears as a promising approach for overcoming statistical heterogeneity, in current approaches its update is biased towards the most recently sampled clients. As we show in this work, this is the reason why it fails to outperform FedAvg, preventing its effective use in real-world large-scale scenarios. In this work, we propose a novel Generalized Heavy-Ball Momentum (GHBM) and theoretically prove it enables convergence under unbounded data heterogeneity in cyclic partial participation, thereby advancing the understanding of momentum's effectiveness in FL. We then introduce adaptive and communication-efficient variants of GHBM that match the communication complexity of FedAvg in settings where clients can be stateful. Extensive experiments on vision and language tasks confirm our theoretical findings, demonstrating that GHBM substantially improves state-of-the-art performance under random uniform client sampling, particularly in large-scale settings with high data heterogeneity and low client participation. Code is available at https://rickzack.github.io/GHBM.
Authors: Saqib Javed, Hieu Le, Mathieu Salzmann
Abstract: A key challenge in Domain Generalization (DG) is preventing overfitting to source domains, which can be mitigated by finding flatter minima in the loss landscape. In this work, we propose Quantization-aware Training for Domain Generalization (QT-DoG) and demonstrate that weight quantization effectively leads to flatter minima in the loss landscape, thereby enhancing domain generalization. Unlike traditional quantization methods focused on model compression, QT-DoG exploits quantization as an implicit regularizer by inducing noise in model weights, guiding the optimization process toward flatter minima that are less sensitive to perturbations and overfitting. We provide both an analytical perspective and empirical evidence demonstrating that quantization inherently encourages flatter minima, leading to better generalization across domains. Moreover, with the benefit of reducing the model size through quantization, we demonstrate that an ensemble of multiple quantized models further yields superior accuracy than the state-of-the-art DG approaches with no computational or memory overheads. Code is released at: https://saqibjaved1.github.io/QT_DoG/.
Authors: Lorenzo Aloisi, Luigi Sigillo, Aurelio Uncini, Danilo Comminiello
Abstract: In recent years, diffusion models have emerged as a superior alternative to generative adversarial networks (GANs) for high-fidelity image generation, with wide applications in text-to-image generation, image-to-image translation, and super-resolution. However, their real-time feasibility is hindered by slow training and inference speeds. This study addresses this challenge by proposing a wavelet-based conditional Diffusion GAN scheme for Single-Image Super-Resolution (SISR). Our approach utilizes the diffusion GAN paradigm to reduce the timesteps required by the reverse diffusion process and the Discrete Wavelet Transform (DWT) to achieve dimensionality reduction, decreasing training and inference times significantly. The results of an experimental validation on the CelebA-HQ dataset confirm the effectiveness of our proposed scheme. Our approach outperforms other state-of-the-art methodologies successfully ensuring high-fidelity output while overcoming inherent drawbacks associated with diffusion models in time-sensitive applications.
Authors: Zi Li, Ying Chen, Zeli Chen, Yanzhou Su, Tai Ma, Tony C. W. Mok, Yan-Jie Zhou, Yunhai Bai, Zhinlin Zheng, Le Lu, Yirui Wang, Jia Ge, Xianghua Ye, Senxiang Yan, Dakai Jin
Abstract: In the radiation therapy of nasopharyngeal carcinoma (NPC), clinicians typically delineate the gross tumor volume (GTV) using non-contrast planning computed tomography to ensure accurate radiation dose delivery. However, the low contrast between tumors and adjacent normal tissues necessitates that radiation oncologists manually delineate the tumors, often relying on diagnostic MRI for guidance. % In this study, we propose a novel approach to directly segment NPC gross tumors on non-contrast planning CT images, circumventing potential registration errors when aligning MRI or MRI-derived tumor masks to planning CT. To address the low contrast issues between tumors and adjacent normal structures in planning CT, we introduce a 3D Semantic Asymmetry Tumor segmentation (SATs) method. Specifically, we posit that a healthy nasopharyngeal region is characteristically bilaterally symmetric, whereas the emergence of nasopharyngeal carcinoma disrupts this symmetry. Then, we propose a Siamese contrastive learning segmentation framework that minimizes the voxel-wise distance between original and flipped areas without tumor and encourages a larger distance between original and flipped areas with tumor. Thus, our approach enhances the sensitivity of features to semantic asymmetries. % Extensive experiments demonstrate that the proposed SATs achieves the leading NPC GTV segmentation performance in both internal and external testing, \emph{e.g.}, with at least 2\% absolute Dice score improvement and 12\% average distance error reduction when compared to other state-of-the-art methods in the external testing.
Authors: Qiushi Sun, Kanzhi Cheng, Zichen Ding, Chuanyang Jin, Yian Wang, Fangzhi Xu, Zhenyu Wu, Chengyou Jia, Liheng Chen, Zhoumianze Liu, Ben Kao, Guohao Li, Junxian He, Yu Qiao, Zhiyong Wu
Abstract: Graphical User Interface (GUI) agents powered by Vision-Language Models (VLMs) have demonstrated human-like computer control capability. Despite their utility in advancing digital automation, a critical bottleneck persists: collecting high-quality trajectory data for training. Common practices for collecting such data rely on human supervision or synthetic data generation through executing pre-defined tasks, which are either resource-intensive or unable to guarantee data quality. Moreover, these methods suffer from limited data diversity and significant gaps between synthetic data and real-world environments. To address these challenges, we propose OS-Genesis, a novel GUI data synthesis pipeline that reverses the conventional trajectory collection process. Instead of relying on pre-defined tasks, OS-Genesis enables agents first to perceive environments and perform step-wise interactions, then retrospectively derive high-quality tasks to enable trajectory-level exploration. A trajectory reward model is then employed to ensure the quality of the generated trajectories. We demonstrate that training GUI agents with OS-Genesis significantly improves their performance on highly challenging online benchmarks. In-depth analysis further validates OS-Genesis's efficiency and its superior data quality and diversity compared to existing synthesis methods. Our codes, data, and checkpoints are available at https://qiushisun.github.io/OS-Genesis-Home/.
Authors: Young Seok Jeon, Hongfei Yang, Huazhu Fu, Mengling Feng
Abstract: 3D models surpass 2D models in CT/MRI segmentation by effectively capturing inter-slice relationships. However, the added depth dimension substantially increases memory consumption. While patch-based training alleviates memory constraints, it significantly slows down the inference speed due to the sliding window (SW) approach. We propose No-More-Sliding-Window (NMSW), a novel end-to-end trainable framework that enhances the efficiency of generic 3D segmentation backbone during an inference step by eliminating the need for SW. NMSW employs a differentiable Top-k module to selectively sample only the most relevant patches, thereby minimizing redundant computations. When patch-level predictions are insufficient, the framework intelligently leverages coarse global predictions to refine results. Evaluated across 3 tasks using 3 segmentation backbones, NMSW achieves competitive accuracy compared to SW inference while significantly reducing computational complexity by 91% (88.0 to 8.00 TMACs). Moreover, it delivers a 9.1x faster inference on the H100 GPU (99.0 to 8.3 sec) and a 11.1x faster inference on the Xeon Gold CPU (2110 to 189 sec). NMSW is model-agnostic, further boosting efficiency when integrated with any existing efficient segmentation backbones. The code is avaialble: https://github.com/Youngseok0001/open_nmsw.
Authors: Zanting Ye, Xiaolong Niu, Xu Han, Xuanbin Wu, Wantong Lu, Yijun Lu, Hao Sun, Yanchao Huang, Hubing Wu, Lijun Lu
Abstract: Organ segmentation in Positron Emission Tomography (PET) plays a vital role in cancer quantification. Low-dose PET (LDPET) provides a safer alternative by reducing radiation exposure. However, the inherent noise and blurred boundaries make organ segmentation more challenging. Additionally, existing PET organ segmentation methods rely on coregistered Computed Tomography (CT) annotations, overlooking the problem of modality mismatch. In this study, we propose LDOS, a novel CT-free ultra-LDPET organ segmentation pipeline. Inspired by Masked Autoencoders (MAE), we reinterpret LDPET as a naturally masked version of Full-Dose PET (FDPET). LDOS adopts a simple yet effective architecture: a shared encoder extracts generalized features, while task-specific decoders independently refine outputs for denoising and segmentation. By integrating CT-derived organ annotations into the denoising process, LDOS improves anatomical boundary recognition and alleviates the PET/CT misalignments. Experiments demonstrate that LDOS achieves state-of-the-art performance with mean Dice scores of 73.11% (18F-FDG) and 73.97% (68Ga-FAPI) across 18 organs in 5% dose PET. Our code will be available at https://github.com/yezanting/LDOS.
Authors: Shuai Shen, Wanhua Li, Yunpeng Zhang, Yap-Peng Tan, Jiwen Lu
Abstract: Talking head synthesis has emerged as a prominent research topic in computer graphics and multimedia, yet most existing methods often struggle to strike a balance between generation quality and computational efficiency, particularly under real-time constraints. In this paper, we propose a novel framework that integrates Gaussian Splatting with a structured Audio Factorization Plane (Audio-Plane) to enable high-quality, audio-synchronized, and real-time talking head generation. For modeling a dynamic talking head, a 4D volume representation, which consists of three axes in 3D space and one temporal axis aligned with audio progression, is typically required. However, directly storing and processing a dense 4D grid is impractical due to the high memory and computation cost, and lack of scalability for longer durations. We address this challenge by decomposing the 4D volume representation into a set of audio-independent spatial planes and audio-dependent planes, forming a compact and interpretable representation for talking head modeling that we refer to as the Audio-Plane. This factorized design allows for efficient and fine-grained audio-aware spatial encoding, and significantly enhances the model's ability to capture complex lip dynamics driven by speech signals. To further improve region-specific motion modeling, we introduce an audio-guided saliency splatting mechanism based on region-aware modulation, which adaptively emphasizes highly dynamic regions such as the mouth area. This allows the model to focus its learning capacity on where it matters most for accurate speech-driven animation. Extensive experiments on both the self-driven and the cross-driven settings demonstrate that our method achieves state-of-the-art visual quality, precise audio-lip synchronization, and real-time performance, outperforming prior approaches across both 2D- and 3D-based paradigms.
Authors: Younghyun Kim, Jongheon Jeong, Sangkyung Kwak, Kyungmin Lee, Juho Lee, Jinwoo Shin
Abstract: Learning robust representations from data often requires scale, which has led to the success of recent zero-shot models such as CLIP. However, the obtained robustness can easily be deteriorated when these models are fine-tuned on other downstream tasks (e.g., of smaller scales). Previous works often interpret this phenomenon in the context of domain shift, developing fine-tuning methods that aim to preserve the original domain as much as possible. However, in a different context, fine-tuned models with limited data are also prone to learning features that are spurious to humans, such as background or texture. In this paper, we propose StarFT (Spurious Textual Alignment Regularization), a novel framework for fine-tuning zero-shot models to enhance robustness by preventing them from learning spuriosity. We introduce a regularization that aligns the output distribution for spuriosity-injected labels with the original zero-shot model, ensuring that the model is not induced to extract irrelevant features further from these descriptions. We leverage recent language models to get such spuriosity-injected labels by generating alternative textual descriptions that highlight potentially confounding features. Extensive experiments validate the robust generalization of StarFT and its emerging properties: zero-shot group robustness and improved zero-shot classification. Notably, StarFT boosts both worst-group and average accuracy by 14.30% and 3.02%, respectively, in the Waterbirds group shift scenario, where other robust fine-tuning baselines show even degraded performance.
Authors: Qiushi Sun, Zhoumianze Liu, Chang Ma, Zichen Ding, Fangzhi Xu, Zhangyue Yin, Haiteng Zhao, Zhenyu Wu, Kanzhi Cheng, Zhaoyang Liu, Jianing Wang, Qintong Li, Xiangru Tang, Tianbao Xie, Xiachong Feng, Xiang Li, Ben Kao, Wenhai Wang, Biqing Qi, Lingpeng Kong, Zhiyong Wu
Abstract: Large Language Models (LLMs) have extended their impact beyond Natural Language Processing, substantially fostering the development of interdisciplinary research. Recently, various LLM-based agents have been developed to assist scientific discovery progress across multiple aspects and domains. Among these, computer-using agents, capable of interacting with operating systems as humans do, are paving the way to automated scientific problem-solving and addressing routines in researchers' workflows. Recognizing the transformative potential of these agents, we introduce ScienceBoard, which encompasses two complementary contributions: (i) a realistic, multi-domain environment featuring dynamic and visually rich scientific workflows with integrated professional software, where agents can autonomously interact via different interfaces to accelerate complex research tasks and experiments; and (ii) a challenging benchmark of 169 high-quality, rigorously validated real-world tasks curated by humans, spanning scientific-discovery workflows in domains such as biochemistry, astronomy, and geoinformatics. Extensive evaluations of agents with state-of-the-art backbones (e.g., GPT-4o, Claude 3.7, UI-TARS) show that, despite some promising results, they still fall short of reliably assisting scientists in complex workflows, achieving only a 15% overall success rate. In-depth analysis further provides valuable insights for addressing current agent limitations and more effective design principles, paving the way to build more capable agents for scientific discovery. Our code, environment, and benchmark are at https://qiushisun.github.io/ScienceBoard-Home/.
Authors: Ren\'e Peinl, Vincent Tischler
Abstract: This paper introduces a novel benchmark dataset designed to evaluate the capabilities of Vision Language Models (VLMs) on tasks that combine visual reasoning with subject-specific background knowledge in the German language. In contrast to widely used English-language benchmarks that often rely on artificially difficult or decontextualized problems, this dataset draws from real middle school curricula across nine domains including mathematics, history, biology, and religion. The benchmark includes over 2,000 open-ended questions grounded in 486 images, ensuring that models must integrate visual interpretation with factual reasoning rather than rely on superficial textual cues. We evaluate thirteen state-of-the-art open-weight VLMs across multiple dimensions, including domain-specific accuracy and performance on adversarial crafted questions. Our findings reveal that even the strongest models achieve less than 45% overall accuracy, with particularly poor performance in music, mathematics, and adversarial settings. Furthermore, the results indicate significant discrepancies between success on popular benchmarks and real-world multimodal understanding. We conclude that middle school-level tasks offer a meaningful and underutilized avenue for stress-testing VLMs, especially in non-English contexts. The dataset and evaluation protocol serve as a rigorous testbed to better understand and improve the visual and linguistic reasoning capabilities of future AI systems.
Authors: Jiayi Zheng, Xiaodong Cun
Abstract: We propose FairyGen, an automatic system for generating story-driven cartoon videos from a single child's drawing, while faithfully preserving its unique artistic style. Unlike previous storytelling methods that primarily focus on character consistency and basic motion, FairyGen explicitly disentangles character modeling from stylized background generation and incorporates cinematic shot design to support expressive and coherent storytelling. Given a single character sketch, we first employ an MLLM to generate a structured storyboard with shot-level descriptions that specify environment settings, character actions, and camera perspectives. To ensure visual consistency, we introduce a style propagation adapter that captures the character's visual style and applies it to the background, faithfully retaining the character's full visual identity while synthesizing style-consistent scenes. A shot design module further enhances visual diversity and cinematic quality through frame cropping and multi-view synthesis based on the storyboard. To animate the story, we reconstruct a 3D proxy of the character to derive physically plausible motion sequences, which are then used to fine-tune an MMDiT-based image-to-video diffusion model. We further propose a two-stage motion customization adapter: the first stage learns appearance features from temporally unordered frames, disentangling identity from motion; the second stage models temporal dynamics using a timestep-shift strategy with frozen identity weights. Once trained, FairyGen directly renders diverse and coherent video scenes aligned with the storyboard. Extensive experiments demonstrate that our system produces animations that are stylistically faithful, narratively structured natural motion, highlighting its potential for personalized and engaging story animation. The code will be available at https://github.com/GVCLab/FairyGen