Peer Revenue Past citations are no guarantee of future returns

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

Computer Science Computer Vision

cs.CV Watch Computer Vision (sign in to use)

Thursday 4 December

Computer Vision

TARDis: Time Attenuated Representation Disentanglement for Incomplete Multi-Modal Tumor Segmentation and Classification

The accurate diagnosis and segmentation of tumors in contrast-enhanced Computed Tomography (CT) are fundamentally driven by the distinctive hemodynamic profiles of contrast agents over time. However, in real-world clinical practice, complete temporal dynamics are often hard to capture by strict radiation dose limits and inconsistent acquisition protocols across institutions, leading to a prevalent missing modality problem. Existing deep learning approaches typically treat missing phases as absent independent channels, ignoring the inherent temporal continuity of hemodynamics. In this work, we propose Time Attenuated Representation Disentanglement (TARDis), a novel physics-aware framework that redefines missing modalities as missing sample points on a continuous Time-Attenuation Curve. We first hypothesize that the latent feature can be disentangled into a time-invariant static component (anatomy) and a time-dependent dynamic component (perfusion). We achieve this via a dual-path architecture: a quantization-based path using a learnable embedding dictionary to extract consistent anatomical structures, and a probabilistic path using a Hemodynamic Conditional Variational Autoencoder to model dynamic enhancement conditioned on the estimated scan time. This design allows the network to infer missing hemodynamic features by sampling from the learned latent distribution. Extensive experiments on a large-scale multi-modal private abdominal CT dataset (2,282 patients) and two public datasets demonstrate that TARDis significantly outperforms state-of-the-art incomplete modality frameworks. Notably, our method maintains robust diagnostic performance even in extreme data-sparsity scenarios, highlighting its potential for reducing radiation exposure while maintaining diagnostic precision.
Read the full abstract

All authors Zishuo Wan, Qinqin Kang, Na Li, Yi Huang, Qianru Zhang, Le Lu, Yun Bian, Dawei Ding, Ke Yan

Computer Vision

Prompt2Craft: Generating Functional Craft Assemblies with LLMs

Inspired by traditional handmade crafts, where a person improvises assemblies based on the available objects, we formally introduce the Craft Assembly Task. It is a robotic assembly task that involves building an accurate representation of a given target object using the available objects, which do not directly correspond to its parts. In this work, we focus on selecting the subset of available objects for the final craft, when the given input is an RGB image of the target in the wild. We use a mask segmentation neural network to identify visible parts, followed by retrieving labeled template meshes. These meshes undergo pose optimization to determine the most suitable template. Then, we propose to simplify the parts of the transformed template mesh to primitive shapes like cuboids or cylinders. Finally, we design a search algorithm to find correspondences in the scene based on local and global proportions. We develop baselines for comparison that consider all possible combinations, and choose the highest scoring combination for common metrics used in foreground maps and mask accuracy. Our approach achieves comparable results to the baselines for two different scenes, and we show qualitative results for an implementation in a real-world scenario.
Read the full abstract

All authors Vitor Hideyo Isume, Takuya Kiyokawa, Natsuki Yamanobe, Yukiyasu Domae, Weiwei Wan, Kensuke Harada

Computer Vision

Dataset creation for supervised deep learning-based analysis of microscopic images – review of important considerations and recommendations

Supervised deep learning (DL) receives great interest for automated analysis of microscopic images with an increasing body of literature supporting its potential. The development and validation of those DL models relies heavily on the availability of high-quality, large-scale datasets. However, creating such datasets is a complex and resource-intensive process, often hindered by challenges such as time constraints, domain variability, and risks of bias in image collection and label creation. This review provides a comprehensive guide to the critical steps in dataset creation, including: 1) image acquisition, 2) selection of annotation software, and 3) annotation creation. In addition to ensuring a sufficiently large number of images, it is crucial to address sources of image variability (domain shifts) - such as those related to slide preparation and digitization - that could lead to algorithmic errors if not adequately represented in the training data. Key quality criteria for annotations are the three "C"s: correctness, completeness, and consistency. This review explores methods to enhance annotation quality through the use of advanced techniques that mitigate the limitations of single annotators. To support dataset creators, a standard operating procedure (SOP) is provided as supplemental material, outlining best practices for dataset development. Furthermore, the article underscores the importance of open datasets in driving innovation and enhancing reproducibility of DL research. By addressing the challenges and offering practical recommendations, this review aims to advance the creation of and availability to high-quality, large-scale datasets, ultimately contributing to the development of generalizable and robust DL models for pathology applications.
Read the full abstract

All authors Christof A. Bertram, Viktoria Weiss, Jonas Ammeling, F. Maria Schabel, Taryn A. Donovan, Frauke Wilm, Christian Marzahl, Katharina Breininger, Marc Aubreville

Computer Vision

COOPER: A Unified Model for Cooperative Perception and Reasoning in Spatial Intelligence

Visual Spatial Reasoning is crucial for enabling Multimodal Large Language Models (MLLMs) to understand object properties and spatial relationships, yet current models still struggle with 3D-aware reasoning. Existing approaches typically enhance either perception, by augmenting RGB inputs with auxiliary modalities such as depth and segmentation, or reasoning, by training on spatial VQA datasets and applying reinforcement learning, and thus treat these two aspects in isolation. In this work, we investigate whether a unified MLLM can develop an intrinsic ability to enhance spatial perception and, through adaptive interleaved reasoning, achieve stronger spatial intelligence. We propose COOPER, a unified MLLM that leverages depth and segmentation as auxiliary modalities and is trained in two stages to acquire auxiliary modality generation and adaptive, interleaved reasoning capabilities. COOPER achieves an average 6.91% improvement in spatial reasoning while maintaining general performance. Moreover, even a variant trained only for auxiliary modality generation attains a 7.92% gain on distance and size estimation, suggesting that learning to generate auxiliary modalities helps internalize spatial knowledge and strengthen spatial understanding.
Read the full abstract

All authors Zefeng Zhang, Xiangzhao Hao, Hengzhu Tang, Zhenyu Zhang, Jiawei Sheng, Xiaodong Li, Zhenyang Li, Li Gao, Daiting Shi, Dawei Yin, Tingwen Liu

Computer Vision · Cross-listed from cs.GR

DISK: Differentiable Sparse Kernel Complex for Efficient Spatially-Variant Convolution

Image convolution with complex kernels is a fundamental operation in photography, scientific imaging, and animation effects, yet direct dense convolution is computationally prohibitive on resource-limited devices. Existing approximations, such as simulated annealing or low-rank decompositions, either lack efficiency or fail to capture non-convex kernels. We introduce a differentiable kernel decomposition framework that represents a target spatially-variant, dense, complex kernel using a set of sparse kernel samples. Our approach features (i) a decomposition that enables differentiable optimization of sparse kernels, (ii) a dedicated initialization strategy for non-convex shapes to avoid poor local minima, and (iii) a kernel-space interpolation scheme that extends single-kernel filtering to spatially varying filtering without retraining and additional runtime overhead. Experiments on Gaussian and non-convex kernels show that our method achieves higher fidelity than simulated annealing and significantly lower cost than low-rank decompositions. Our approach provides a practical solution for mobile imaging and real-time rendering, while remaining fully differentiable for integration into broader learning pipelines.
Read the full abstract

Computer Vision

Counterfeit Answers: Adversarial Forgery against OCR-Free Document Visual Question Answering

Document Visual Question Answering (DocVQA) enables end-to-end reasoning grounded on information present in a document input. While recent models have shown impressive capabilities, they remain vulnerable to adversarial attacks. In this work, we introduce a novel attack scenario that aims to forge document content in a visually imperceptible yet semantically targeted manner, allowing an adversary to induce specific or generally incorrect answers from a DocVQA model. We develop specialized attack algorithms that can produce adversarially forged documents tailored to different attackers' goals, ranging from targeted misinformation to systematic model failure scenarios. We demonstrate the effectiveness of our approach against two end-to-end state-of-the-art models: Pix2Struct, a vision-language transformer that jointly processes image and text through sequence-to-sequence modeling, and Donut, a transformer-based model that directly extracts text and answers questions from document images. Our findings highlight critical vulnerabilities in current DocVQA systems and call for the development of more robust defenses. We release our open source code at https://github.com/pralab/adv-docVQA.
Read the full abstract

All authors Marco Pintore, Maura Pintor, Dimosthenis Karatzas, Battista Biggio

Computer Vision

Gaussian Entropy Fields: Driving Adaptive Sparsity in 3D Gaussian Optimization

3D Gaussian Splatting (3DGS) has emerged as a leading technique for novel view synthesis, demonstrating exceptional rendering efficiency. \replaced[]{Well-reconstructed surfaces can be characterized by low configurational entropy, where dominant primitives clearly define surface geometry while redundant components are suppressed.}{The key insight is that well-reconstructed surfaces naturally exhibit low configurational entropy, where dominant primitives clearly define surface geometry while suppressing redundant components.} Three complementary technical contributions are introduced: (1) entropy-driven surface modeling via entropy minimization for low configurational entropy in primitive distributions; (2) adaptive spatial regularization using the Surface Neighborhood Redundancy Index (SNRI) and image entropy-guided weighting; (3) multi-scale geometric preservation through competitive cross-scale entropy alignment. Extensive experiments demonstrate that GEF achieves competitive geometric precision on DTU and T&T benchmarks, while delivering superior rendering quality compared to existing methods on Mip-NeRF 360. Notably, superior Chamfer Distance (0.64) on DTU and F1 score (0.44) on T&T are obtained, alongside the best SSIM (0.855) and LPIPS (0.136) among baselines on Mip-NeRF 360, validating the framework's ability to enhance surface reconstruction accuracy without compromising photometric fidelity.
Read the full abstract

Computer Vision

VideoMem: Enhancing Ultra-Long Video Understanding via Adaptive Memory Management

Ultra long video understanding remains an open challenge, as existing vision language models (VLMs) falter on such content due to limited context length and inefficient long term memory retention. To address this, recent works have attempted to construct external knowledge bases and corresponding retrieval agumented generation (RAG) systems, yet these incur enormous storage and computational overhead. In this paper, we propose VideoMem, a novel framework that pioneers models long video understanding as a sequential generation task via adaptive memory management. Specifically, VideoMem dynamically updates a global memory buffer, which adaptively retains critical information while discarding redundant content across the video timeline. To efficiently train VLMs for such long-term tasks, VideoMem integrates the Progressive Grouped Relative Policy Optimization (PRPO) algorithm, equipped with two core modules: Progressive State Propagation (PSP) adaptively retains valid current states, propagates them to the next rollout step, and gradually narrows the model exploration space. Temporal Cascading Reward (TCR) further alleviates reward sparsity, improving sample utilization and accelerating convergence. Extensive experiments demonstrate that VideoMem significantly outperforms existing open-source models across diverse benchmarks for ultra-long video understanding tasks.
Read the full abstract

All authors Hongbo Jin, Qingyuan Wang, Wenhao Zhang, Yang Liu, Sijie Cheng

Computer Vision

X-Humanoid: Robotize Human Videos to Generate Humanoid Videos at Scale

The advancement of embodied AI has unlocked significant potential for intelligent humanoid robots. However, progress in both Vision-Language-Action (VLA) models and world models is severely hampered by the scarcity of large-scale, diverse training data. A promising solution is to "robotize" web-scale human videos, which has been proven effective for policy training. However, these solutions mainly "overlay" robot arms to egocentric videos, which cannot handle complex full-body motions and scene occlusions in third-person videos, making them unsuitable for robotizing humans. To bridge this gap, we introduce X-Humanoid, a generative video editing approach that adapts the powerful Wan 2.2 model into a video-to-video structure and finetunes it for the human-to-humanoid translation task. This finetuning requires paired human-humanoid videos, so we designed a scalable data creation pipeline, turning community assets into 17+ hours of paired synthetic videos using Unreal Engine. We then apply our trained model to 60 hours of the Ego-Exo4D videos, generating and releasing a new large-scale dataset of over 3.6 million "robotized" humanoid video frames. Quantitative analysis and user studies confirm our method's superiority over existing baselines: 69% of users rated it best for motion consistency, and 62.1% for embodiment correctness.
Read the full abstract

All authors Pei Yang, Hai Ci, Yiren Song, Mike Zheng Shou

Computer Vision

Detection of Intoxicated Individuals from Facial Video Sequences via a Recurrent Fusion Model

Alcohol consumption is a significant public health concern and a major cause of accidents and fatalities worldwide. This study introduces a novel video-based facial sequence analysis approach dedicated to the detection of alcohol intoxication. The method integrates facial landmark analysis via a Graph Attention Network (GAT) with spatiotemporal visual features extracted using a 3D ResNet. These features are dynamically fused with adaptive prioritization to enhance classification performance. Additionally, we introduce a curated dataset comprising 3,542 video segments derived from 202 individuals to support training and evaluation. Our model is compared against two baselines: a custom 3D-CNN and a VGGFace+LSTM architecture. Experimental results show that our approach achieves 95.82% accuracy, 0.977 precision, and 0.97 recall, outperforming prior methods. The findings demonstrate the model's potential for practical deployment in public safety systems for non-invasive, reliable alcohol intoxication detection.
Read the full abstract

Computer Vision

Refaçade: Editing Object with Given Reference Texture

Recent advances in diffusion models have brought remarkable progress in image and video editing, yet some tasks remain underexplored. In this paper, we introduce a new task, Object Retexture, which transfers local textures from a reference object to a target object in images or videos. To perform this task, a straightforward solution is to use ControlNet conditioned on the source structure and the reference texture. However, this approach suffers from limited controllability for two reasons: conditioning on the raw reference image introduces unwanted structural information, and it fails to disentangle the visual texture and structure information of the source. To address this problem, we propose Refaçade, a method that consists of two key designs to achieve precise and controllable texture transfer in both images and videos. First, we employ a texture remover trained on paired textured/untextured 3D mesh renderings to remove appearance information while preserving the geometry and motion of source videos. Second, we disrupt the reference global layout using a jigsaw permutation, encouraging the model to focus on local texture statistics rather than the global layout of the object. Extensive experiments demonstrate superior visual quality, precise editing, and controllability, outperforming strong baselines in both quantitative and human evaluations. Code is available at https://github.com/fishZe233/Refacade.
Read the full abstract

All authors Youze Huang, Penghui Ruan, Bojia Zi, Xianbiao Qi, Jianan Wang, Rong Xiao

Computer Vision

PhyVLLM: Physics-Guided Video Language Model with Motion-Appearance Disentanglement

Video Large Language Models (Video LLMs) have shown impressive performance across a wide range of video-language tasks. However, they often fail in scenarios requiring a deeper understanding of physical dynamics. This limitation primarily arises from their reliance on appearance-based matching. Incorporating physical motion modeling is crucial for deeper video understanding, but presents three key challenges: (1) motion signals are often entangled with appearance variations, making it difficult to extract clean physical cues; (2) effective motion modeling requires not only continuous-time motion representations but also capturing physical dynamics; and (3) collecting accurate annotations for physical attributes is costly and often impractical. To address these issues, we propose PhyVLLM, a physical-guided video-language framework that explicitly incorporates physical motion into Video LLMs. Specifically, PhyVLLM disentangles visual appearance and object motion through a dual-branch encoder. To model physical dynamics over time, we incorporate a Neural Ordinary Differential Equation (Neural ODE) module, which generates differentiable physical dynamic representations. The resulting motion-aware representations are projected into the token space of a pretrained LLM, enabling physics reasoning without compromising the model's original multimodal capabilities. To circumvent the need for explicit physical labels, PhyVLLM employs a self-supervised manner to model the continuous evolution of object motion. Experimental results demonstrate that PhyVLLM significantly outperforms state-of-the-art Video LLMs on both physical reasoning and general video understanding tasks, highlighting the advantages of incorporating explicit physical modeling.
Read the full abstract

All authors Yu-Wei Zhan, Xin Wang, Hong Chen, Tongtong Feng, Wei Feng, Ren Wang, Guangyao Li, Qing Li, Wenwu Zhu

Computer Vision

Auto3R: Automated 3D Reconstruction and Scanning via Data-driven Uncertainty Quantification

Traditional high-quality 3D scanning and reconstruction typically relies on human labor to plan the scanning procedure. With the rapid development of embodied systems such as drones and robots, there is a growing demand of performing accurate 3D scanning and reconstruction in an fully automated manner. We introduce Auto3R, a data-driven uncertainty quantification model that is designed to automate the 3D scanning and reconstruction of scenes and objects, including objects with non-lambertian and specular materials. Specifically, in a process of iterative 3D reconstruction and scanning, Auto3R can make efficient and accurate prediction of uncertainty distribution over potential scanning viewpoints, without knowing the ground truth geometry and appearance. Through extensive experiments, Auto3R achieves superior performance that outperforms the state-of-the-art methods by a large margin. We also deploy Auto3R on a robot arm equipped with a camera and demonstrate that Auto3R can be used to effectively digitize real-world 3D objects and delivers ready-to-use and photorealistic digital assets. Our homepage: https://tomatoma00.github.io/auto3r.github.io .
Read the full abstract

All authors Chentao Shen, Sizhe Zheng, Bingqian Wu, Yaohua Feng, Yuanchen Fei, Mingyu Mei, Hanwen Jiang, Xiangru Huang

Computer Vision

Identity Clue Refinement and Enhancement for Visible-Infrared Person Re-Identification

Visible-Infrared Person Re-Identification (VI-ReID) is a challenging cross-modal matching task due to significant modality discrepancies. While current methods mainly focus on learning modality-invariant features through unified embedding spaces, they often focus solely on the common discriminative semantics across modalities while disregarding the critical role of modality-specific identity-aware knowledge in discriminative feature learning. To bridge this gap, we propose a novel Identity Clue Refinement and Enhancement (ICRE) network to mine and utilize the implicit discriminative knowledge inherent in modality-specific attributes. Initially, we design a Multi-Perception Feature Refinement (MPFR) module that aggregates shallow features from shared branches, aiming to capture modality-specific attributes that are easily overlooked. Then, we propose a Semantic Distillation Cascade Enhancement (SDCE) module, which distills identity-aware knowledge from the aggregated shallow features and guide the learning of modality-invariant features. Finally, an Identity Clues Guided (ICG) Loss is proposed to alleviate the modality discrepancies within the enhanced features and promote the learning of a diverse representation space. Extensive experiments across multiple public datasets clearly show that our proposed ICRE outperforms existing SOTA methods.
Read the full abstract

All authors Guoqing Zhang, Zhun Wang, Hairui Wang, Zhonglin Ye, Yuhui Zheng

Computer Vision

WiFi-based Cross-Domain Gesture Recognition Using Attention Mechanism

While fulfilling communication tasks, wireless signals can also be used to sense the environment. Among various types of sensing media, WiFi signals offer advantages such as widespread availability, low hardware cost, and strong robustness to environmental conditions like light, temperature, and humidity. By analyzing Wi-Fi signals in the environment, it is possible to capture dynamic changes of the human body and accomplish sensing applications such as gesture recognition. Although many existing gesture sensing solutions perform well in-domain but lack cross-domain capabilities (i.e., recognition performance in untrained environments). To address this, we extract Doppler spectra from the channel state information (CSI) received by all receivers and concatenate each Doppler spectrum along the same time axis to generate fused images with multi-angle information as input features. Furthermore, inspired by the convolutional block attention module (CBAM), we propose a gesture recognition network that integrates a multi-semantic spatial attention mechanism with a self-attention-based channel mechanism. This network constructs attention maps to quantify the spatiotemporal features of gestures in images, enabling the extraction of key domain-independent features. Additionally, ResNet18 is employed as the backbone network to further capture deep-level features. To validate the network performance, we evaluate the proposed network on the public Widar3 dataset, and the results show that it not only maintains high in-domain accuracy of 99.72%, but also achieves high performance in cross-domain recognition of 97.61%, significantly outperforming existing best solutions.
Read the full abstract

All authors Ruijing Liu, Cunhua Pan, Jiaming Zeng, Hong Ren, Kezhi Wang, Lei Kong, Jiangzhou Wang

Computer Vision

Boundary-Aware Test-Time Adaptation for Zero-Shot Medical Image Segmentation

Due to the scarcity of annotated data and the substantial computational costs of model, conventional tuning methods in medical image segmentation face critical challenges. Current approaches to adapting pretrained models, including full-parameter and parameter-efficient fine-tuning, still rely heavily on task-specific training on downstream tasks. Therefore, zero-shot segmentation has gained increasing attention, especially with foundation models such as SAM demonstrating promising generalization capabilities. However, SAM still faces notable limitations on medical datasets due to domain shifts, making efficient zero-shot enhancement an urgent research goal. To address these challenges, we propose BA-TTA-SAM, a task-agnostic test-time adaptation framework that significantly enhances the zero-shot segmentation performance of SAM via test-time adaptation. This framework integrates two key mechanisms: (1) The encoder-level Gaussian prompt injection embeds Gaussian-based prompts directly into the image encoder, providing explicit guidance for initial representation learning. (2) The cross-layer boundary-aware attention alignment exploits the hierarchical feature interactions within the ViT backbone, aligning deep semantic responses with shallow boundary cues. Experiments on four datasets, including ISIC, Kvasir, BUSI, and REFUGE, show an average improvement of 12.4% in the DICE score compared with SAM's zero-shot segmentation performance. The results demonstrate that our method consistently outperforms state-of-the-art models in medical image segmentation. Our framework significantly enhances the generalization ability of SAM, without requiring any source-domain training data. Extensive experiments on publicly available medical datasets strongly demonstrate the superiority of our framework. Our code is available at https://github.com/Emilychenlin/BA-TTA-SAM.
Read the full abstract

All authors Chenlin Xu, Lei Zhang, Lituan Wang, Xinyu Pu, Pengfei Ma, Guangwu Qian, Zizhou Wang, Yan Wang

Computer Vision

VideoSSM: Autoregressive Long Video Generation with Hybrid State-Space Memory

Autoregressive (AR) diffusion enables streaming, interactive long-video generation by producing frames causally, yet maintaining coherence over minute-scale horizons remains challenging due to accumulated errors, motion drift, and content repetition. We approach this problem from a memory perspective, treating video synthesis as a recurrent dynamical process that requires coordinated short- and long-term context. We propose VideoSSM, a Long Video Model that unifies AR diffusion with a hybrid state-space memory. The state-space model (SSM) serves as an evolving global memory of scene dynamics across the entire sequence, while a context window provides local memory for motion cues and fine details. This hybrid design preserves global consistency without frozen, repetitive patterns, supports prompt-adaptive interaction, and scales in linear time with sequence length. Experiments on short- and long-range benchmarks demonstrate state-of-the-art temporal consistency and motion stability among autoregressive video generator especially at minute-scale horizons, enabling content diversity and interactive prompt-based control, thereby establishing a scalable, memory-aware framework for long video generation.
Read the full abstract

All authors Yifei Yu, Xiaoshan Wu, Xinting Hu, Tao Hu, Yangtian Sun, Xiaoyang Lyu, Bo Wang, Lin Ma, Yuewen Ma, Zhongrui Wang, Xiaojuan Qi

Computer Vision

EgoLCD: Egocentric Video Generation with Long Context Diffusion

Generating long, coherent egocentric videos is difficult, as hand-object interactions and procedural tasks require reliable long-term memory. Existing autoregressive models suffer from content drift, where object identity and scene semantics degrade over time. To address this challenge, we introduce EgoLCD, an end-to-end framework for egocentric long-context video generation that treats long video synthesis as a problem of efficient and stable memory management. EgoLCD combines a Long-Term Sparse KV Cache for stable global context with an attention-based short-term memory, extended by LoRA for local adaptation. A Memory Regulation Loss enforces consistent memory usage, and Structured Narrative Prompting provides explicit temporal guidance. Extensive experiments on the EgoVid-5M benchmark demonstrate that EgoLCD achieves state-of-the-art performance in both perceptual quality and temporal consistency, effectively mitigating generative forgetting and representing a significant step toward building scalable world models for embodied AI. Code: https://github.com/AIGeeksGroup/EgoLCD. Website: https://aigeeksgroup.github.io/EgoLCD.
Read the full abstract

All authors Liuzhou Zhang, Jiarui Ye, Yuanlei Wang, Ming Zhong, Mingju Cao, Wanke Xia, Bowen Zeng, Zeyu Zhang, Hao Tang

Computer Vision

DuGI-MAE: Improving Infrared Mask Autoencoders via Dual-Domain Guidance

Infrared imaging plays a critical role in low-light and adverse weather conditions. However, due to the distinct characteristics of infrared images, existing foundation models such as Masked Autoencoder (MAE) trained on visible data perform suboptimal in infrared image interpretation tasks. To bridge this gap, an infrared foundation model known as InfMAE was developed and pre-trained on large-scale infrared datasets. Despite its effectiveness, InfMAE still faces several limitations, including the omission of informative tokens, insufficient modeling of global associations, and neglect of non-uniform noise. In this paper, we propose a Dual-domain Guided Infrared foundation model based on MAE (DuGI-MAE). First, we design a deterministic masking strategy based on token entropy, preserving only high-entropy tokens for reconstruction to enhance informativeness. Next, we introduce a Dual-Domain Guidance (DDG) module, which simultaneously captures global token relationships and adaptively filters non-uniform background noise commonly present in infrared imagery. To facilitate large-scale pretraining, we construct Inf-590K, a comprehensive infrared image dataset encompassing diverse scenes, various target types, and multiple spatial resolutions. Pretrained on Inf-590K, DuGI-MAE demonstrates strong generalization capabilities across various downstream tasks, including infrared object detection, semantic segmentation, and small target detection. Experimental results validate the superiority of the proposed method over both supervised and self-supervised comparison methods. Our code is available in the supplementary material.
Read the full abstract

All authors Yinghui Xing, Xiaoting Su, Shizhou Zhang, Donghao Chu, Di Xu

Computer Vision

UltraImage: Rethinking Resolution Extrapolation in Image Diffusion Transformers

Recent image diffusion transformers achieve high-fidelity generation, but struggle to generate images beyond these scales, suffering from content repetition and quality degradation. In this work, we present UltraImage, a principled framework that addresses both issues. Through frequency-wise analysis of positional embeddings, we identify that repetition arises from the periodicity of the dominant frequency, whose period aligns with the training resolution. We introduce a recursive dominant frequency correction to constrain it within a single period after extrapolation. Furthermore, we find that quality degradation stems from diluted attention and thus propose entropy-guided adaptive attention concentration, which assigns higher focus factors to sharpen local attention for fine detail and lower ones to global attention patterns to preserve structural consistency. Experiments show that UltraImage consistently outperforms prior methods on Qwen-Image and Flux (around 4K) across three generation scenarios, reducing repetition and improving visual fidelity. Moreover, UltraImage can generate images up to 6K*6K without low-resolution guidance from a training resolution of 1328p, demonstrating its extreme extrapolation capability. Project page is available at \href{https://thu-ml.github.io/ultraimage.github.io/}{https://thu-ml.github.io/ultraimage.github.io/}.
Read the full abstract

All authors Min Zhao, Bokai Yan, Xue Yang, Hongzhou Zhu, Jintao Zhang, Shilong Liu, Chongxuan Li, Jun Zhu

Computer Vision

Back to Basics: Motion Representation Matters for Human Motion Generation Using Diffusion Model

Diffusion models have emerged as a widely utilized and successful methodology in human motion synthesis. Task-oriented diffusion models have significantly advanced action-to-motion, text-to-motion, and audio-to-motion applications. In this paper, we investigate fundamental questions regarding motion representations and loss functions in a controlled study, and we enumerate the impacts of various decisions in the workflow of the generative motion diffusion model. To answer these questions, we conduct empirical studies based on a proxy motion diffusion model (MDM). We apply v loss as the prediction objective on MDM (vMDM), where v is the weighted sum of motion data and noise. We aim to enhance the understanding of latent data distributions and provide a foundation for improving the state of conditional motion diffusion models. First, we evaluate the six common motion representations in the literature and compare their performance in terms of quality and diversity metrics. Second, we compare the training time under various configurations to shed light on how to speed up the training process of motion diffusion models. Finally, we also conduct evaluation analysis on a large motion dataset. The results of our experiments indicate clear performance differences across motion representations in diverse datasets. Our results also demonstrate the impacts of distinct configurations on model training and suggest the importance and effectiveness of these decisions on the outcomes of motion diffusion models.
Read the full abstract

Computer Vision

Shift-Window Meets Dual Attention: A Multi-Model Architecture for Specular Highlight Removal

Inevitable specular highlights in practical environments severely impair the visual performance, thus degrading the task effectiveness and efficiency. Although there exist considerable methods that focus on local information from convolutional neural network models or global information from transformer models, the single-type model falls into a modeling dilemma between local fine-grained details and global long-range dependencies, thus deteriorating for specular highlights with different scales. Therefore, to accommodate specular highlights of all scales, we propose a multi-model architecture for specular highlight removal (MM-SHR) that effectively captures fine-grained features in highlight regions and models long-range dependencies between highlight and highlight-free areas. Specifically, we employ convolution operations to extract local details in the shallow layers of MM-SHR, and utilize the attention mechanism to capture global features in the deep layers, ensuring both operation efficiency and removal accuracy. To model long-range dependencies without compromising computational complexity, we utilize a coarse-to-fine manner and propose Omni-Directional Attention Integration Block(OAIBlock) and Adaptive Region-Aware Hybrid-Domain Dual Attention Convolutional Network(HDDAConv) , which leverage omni-directiona pixel-shifting and window-dividing operations at the raw features to achieve specular highlight removal. Extensive experimental results on three benchmark tasks and six types of surface materials demonstrate that MM-SHR outperforms state-of-the-art methods in both accuracy and efficiency for specular highlight removal. The implementation will be made publicly available at https://github.com/Htcicv/MM-SHR.
Read the full abstract

All authors Tianci Huo, Lingfeng Qi, Yuhan Chen, Qihong Xue, Jinyuan Shao, Hai Yu, Jie Li, Zhanhua Zhang, Guofa Li

Computer Vision

Controllable Long-term Motion Generation with Extended Joint Targets

Generating stable and controllable character motion in real-time is a key challenge in computer animation. Existing methods often fail to provide fine-grained control or suffer from motion degradation over long sequences, limiting their use in interactive applications. We propose COMET, an autoregressive framework that runs in real time, enabling versatile character control and robust long-horizon synthesis. Our efficient Transformer-based conditional VAE allows for precise, interactive control over arbitrary user-specified joints for tasks like goal-reaching and in-betweening from a single model. To ensure long-term temporal stability, we introduce a novel reference-guided feedback mechanism that prevents error accumulation. This mechanism also serves as a plug-and-play stylization module, enabling real-time style transfer. Extensive evaluations demonstrate that COMET robustly generates high-quality motion at real-time speeds, significantly outperforming state-of-the-art approaches in complex motion control tasks and confirming its readiness for demanding interactive applications.
Read the full abstract

All authors Eunjong Lee, Eunhee Kim, Sanghoon Hong, Eunho Jung, Jihoon Kim

Computer Vision

Not All Birds Look The Same: Identity-Preserving Generation For Birds

Since the advent of controllable image generation, increasingly rich modes of control have enabled greater customization and accessibility for everyday users. Zero-shot, identity-preserving models such as Insert Anything and OminiControl now support applications like virtual try-on without requiring additional fine-tuning. While these models may be fitting for humans and rigid everyday objects, they still have limitations for non-rigid or fine-grained categories. These domains often lack accessible, high-quality data – especially videos or multi-view observations of the same subject – making them difficult both to evaluate and to improve upon. Yet, such domains are essential for moving beyond content creation toward applications that demand accuracy and fine detail. Birds are an excellent domain for this task: they exhibit high diversity, require fine-grained cues for identification, and come in a wide variety of poses. We introduce the NABirds Look-Alikes (NABLA) dataset, consisting of 4,759 expert-curated image pairs. Together with 1,073 pairs collected from multi-image observations on iNaturalist and a small set of videos, this forms a benchmark for evaluating identity-preserving generation of birds. We show that state-of-the-art baselines fail to maintain identity on this dataset, and we demonstrate that training on images grouped by species, age, and sex – used as a proxy for identity – substantially improves performance on both seen and unseen species.
Read the full abstract

Computer Vision

DeRA: Decoupled Representation Alignment for Video Tokenization

This paper presents DeRA, a novel 1D video tokenizer that decouples the spatial-temporal representation learning in video tokenization to achieve better training efficiency and performance. Specifically, DeRA maintains a compact 1D latent space while factorizing video encoding into appearance and motion streams, which are aligned with pretrained vision foundation models to capture the spatial semantics and temporal dynamics in videos separately. To address the gradient conflicts introduced by the heterogeneous supervision, we further propose the Symmetric Alignment-Conflict Projection (SACP) module that proactively reformulates gradients by suppressing the components along conflicting directions. Extensive experiments demonstrate that DeRA outperforms LARP, the previous state-of-the-art video tokenizer by 25% on UCF-101 in terms of rFVD. Moreover, using DeRA for autoregressive video generation, we also achieve new state-of-the-art results on both UCF-101 class-conditional generation and K600 frame prediction.
Read the full abstract

All authors Pengbo Guo, Junke Wang, Zhen Xing, Chengxu Liu, Daoguo Dong, Xueming Qian, Zuxuan Wu

Computer Vision · Cross-listed from cs.LG

Feature Engineering vs. Deep Learning for Automated Coin Grading: A Comparative Study on Saint-Gaudens Double Eagles

We challenge the common belief that deep learning always trumps older techniques, using the example of grading Saint-Gaudens Double Eagle gold coins automatically. In our work, we put a feature-based Artificial Neural Network built around 192 custom features pulled from Sobel edge detection and HSV color analysis up against a hybrid Convolutional Neural Network that blends in EfficientNetV2, plus a straightforward Support Vector Machine as the control. Testing 1,785 coins graded by experts, the ANN nailed 86% exact matches and hit 98% when allowing a 3-grade leeway. On the flip side, CNN and SVM mostly just guessed the most common grade, scraping by with 31% and 30% exact hits. Sure, the CNN looked good on broader tolerance metrics, but that is because of some averaging trick in regression that hides how it totally flops at picking out specific grades. All told, when you are stuck with under 2,000 examples and lopsided classes, baking in real coin-expert knowledge through feature design beats out those inscrutable, all-in-one deep learning setups. This rings true for other niche quality checks where data's thin and know-how matters more than raw compute.
Read the full abstract

All authors Tanmay Dogra, Eric Ngo, Mohammad Alam, Jean-Paul Talavera, Asim Dahal

Computer Vision

UniTS: Unified Spatio-Temporal Generative Model for Remote Sensing

One of the primary objectives of satellite remote sensing is to capture the complex dynamics of the Earth environment, which encompasses tasks such as reconstructing continuous cloud-free image sequences, detecting land cover changes, and forecasting future surface evolution. However, existing methods typically require specialized models tailored to different tasks, and lack a general framework that can address these multi-level tasks from a unified perspective. In this paper, we propose a Unified Spatio-Temporal Generative Model (UniTS), which integrates several long-separated core tasks, including time series reconstruction, time series cloud removal, time series semantic change detection, and time series forecasting. Based on the flow matching generative paradigm, UniTS constructs a deterministic evolution path from noise to targets under the guidance of task-specific conditions, achieving unified modeling of spatiotemporal representations for multi-level tasks. The UniTS architecture consists of a diffusion transformer with spatiotemporal blocks, where we design an Adaptive Condition Injector (ACor) to enhance the model's conditional perception of multimodal inputs, enabling high-quality controllable generation. Additionally, we design a Spatiotemporal-aware Modulator (STM) to improve the ability of spatiotemporal blocks to capture complex spatiotemporal dependencies. It substantially outperforms existing specialized models, particularly under challenging conditions such as severe cloud contamination, modality absence, and forecasting complex phenological variations.
Read the full abstract

All authors Yuxiang Zhang, Shunlin Liang, Wenyuan Li, Han Ma, Jianglei Xu, Yichuan Ma, Jiangwei Xie, Wei Li, Mengmeng Zhang, Ran Tao, Xiang-Gen Xia

Computer Vision

dVLM-AD: Enhance Diffusion Vision-Language-Model for Driving via Controllable Reasoning

The autonomous driving community is increasingly focused on addressing the challenges posed by out-of-distribution (OOD) driving scenarios. A dominant research trend seeks to enhance end-to-end (E2E) driving systems by integrating vision-language models (VLMs), leveraging their rich world knowledge and reasoning abilities to improve generalization across diverse environments. However, most existing VLMs or vision-language agents (VLAs) are built upon autoregressive (AR) models. In this paper, we observe that existing AR-based VLMs – limited by causal attention and sequential token generation – often fail to maintain consistency and controllability between high-level reasoning and low-level planning. In contrast, recent discrete diffusion VLMs equipped with bidirectional attention exhibit superior controllability and reliability through iterative denoising. Building on these observations, we introduce dVLM-AD, a diffusion-based vision-language model that unifies perception, structured reasoning, and low-level planning for end-to-end driving. Evaluated on nuScenes and WOD-E2E, dVLM-AD yields more consistent reasoning-action pairs and achieves planning performance comparable to existing driving VLM/VLA systems despite a modest backbone, outperforming AR-based baselines with a 9 percent improvement in behavior-trajectory consistency and a 6 percent increase in RFS on long-tail WOD-E2E scenarios. These results suggest a controllable and reliable pathway for scalable end-to-end driving.
Read the full abstract

All authors Yingzi Ma, Yulong Cao, Wenhao Ding, Shuibai Zhang, Yan Wang, Boris Ivanovic, Ming Jiang, Marco Pavone, Chaowei Xiao

Computer Vision

GuidNoise: Single-Pair Guided Diffusion for Generalized Noise Synthesis

Recent image denoising methods have leveraged generative modeling for real noise synthesis to address the costly acquisition of real-world noisy data. However, these generative models typically require camera metadata and extensive target-specific noisy-clean image pairs, often showing limited generalization between settings. In this paper, to mitigate the prerequisites, we propose a Single-Pair Guided Diffusion for generalized noise synthesis GuidNoise, which uses a single noisy/clean pair as the guidance, often easily obtained by itself within a training set. To train GuidNoise, which generates synthetic noisy images from the guidance, we introduce a guidance-aware affine feature modification (GAFM) and a noise-aware refine loss to leverage the inherent potential of diffusion models. This loss function refines the diffusion model's backward process, making the model more adept at generating realistic noise distributions. The GuidNoise synthesizes high-quality noisy images under diverse noise environments without additional metadata during both training and inference. Additionally, GuidNoise enables the efficient generation of noisy-clean image pairs at inference time, making synthetic noise readily applicable for augmenting training data. This self-augmentation significantly improves denoising performance, especially in practical scenarios with lightweight models and limited training data. The code is available at https://github.com/chjinny/GuidNoise.
Read the full abstract

Computer Vision

StreamEQA: Towards Streaming Video Understanding for Embodied Scenarios

As embodied intelligence advances toward real-world deployment, the ability to continuously perceive and reason over streaming visual inputs becomes essential. In such settings, an agent must maintain situational awareness of its environment, comprehend the interactions with surrounding entities, and dynamically plan actions informed by past observations, current contexts, and anticipated future events. To facilitate progress in this direction, we introduce StreamEQA, the first benchmark designed for streaming video question answering in embodied scenarios. StreamEQA evaluates existing MLLMs along two orthogonal dimensions: Embodied and Streaming. Along the embodied dimension, we categorize the questions into three levels: perception, interaction, and planning, which progressively assess a model's ability to recognize fine-grained visual details, reason about agent-object interactions, and perform high-level goal-directed reasoning. For the streaming dimension, questions are divided into backward, real-time, and forward reasoning, with each mode relying on a distinct temporal context. Built upon 156 independent long videos, StreamEQA defines 42 tasks and generates approximately 21K question-answer pairs with precise timestamps through a hybrid pipeline combining automated generation and human refinement. Evaluations of 13 state-of-the-art video-LLMs reveal that, despite strong performance on conventional benchmarks, these models still struggle with streaming video understanding in embodied scenarios. We hope StreamEQA will catalyze research on streaming video understanding for embodied applications.
Read the full abstract

All authors Yifei Wang, Zhenkai Li, Tianwen Qian, Huanran Zheng, Zheng Wang, Yuqian Fu, Xiaoling Wang