CVPR 2025 Papers

Contrastive decoding strategies are widely used to mitigate object hallucinations in multimodal large language models (MLLMs). By reducing over-reliance on language priors, these strategies ensure that generated content remains closely grounded in visual inputs, producing contextually accurate outputs. Since contrastive decoding requires no additional training or external tools, it offers both computational efficiency and versatility, making it highly attractive. However, these methods present two main limitations: (1) bluntly suppressing language priors can compromise coherence and accuracy of generated content, and (2) processing contrastive inputs adds computational load, significantly slowing inference speed. To address these challenges, we propose Visual Amplification Fusion (VAF), a plug-and-play technique that enhances attention to visual signals within the model's middle layers, where modality fusion predominantly occurs. This approach enables more effective capture of visual features, reducing the model's bias toward language modality. Experimental results demonstrate that VAF significantly reduces hallucinations across various MLLMs without affecting inference speed, while maintaining coherence and accuracy in generated outputs.

Internal features from large-scale pre-trained diffusion models have recently been established as powerful semantic descriptors for a wide range of downstream tasks. Works that use these features generally need to add noise to images before passing them through the model to obtain the semantic features, as the models do not offer the most useful features when given images with little to no noise. We show that this noise has a critical impact on the usefulness of these features that cannot be remedied by ensembling with different random noises. We address this issue by introducing a lightweight, unsupervised fine-tuning method that enables diffusion backbones to provide high-quality, noise-free semantic features. We show that these features readily outperform previous diffusion features by a wide margin in a wide variety of extraction setups and downstream tasks, offering better performance than even ensemble-based methods at a fraction of the cost.

A person downloading a pre-trained model from the web should be aware of its biases. Existing approaches for bias identification rely on datasets containing labels for the task of interest, something that a non-expert may not have access to, or may not have the necessary resources to collect: this greatly limits the number of tasks where model biases can be identified. In this work, we present Classifier-to-Bias (C2B), the first bias discovery framework that works without access to any labeled data: it only relies on a textual description of the classification task to identify biases in the target classification model. This description is fed to a large language model to generate bias proposals and corresponding captions depicting biases together with task-specific target labels. A retrieval model collects images for those captions, which are then used to assess the accuracy of the model w.r.t. the given biases. C2B is training-free, does not require any annotations, has no constraints on the list of biases, and can be applied to any pre-trained model on any classification task. Experiments on two publicly available datasets show that C2B discovers biases beyond those of the original datasets and outperforms a recent state-of-the-art bias detection baseline that relies on task-specific annotations, being a promising first step toward addressing task-agnostic unsupervised bias detection.

Multi-label classification is crucial for comprehensive image understanding, yet acquiring accurate annotations is challenging and costly. To address this, a recent study suggests exploiting unsupervised multi-label classification leveraging CLIP, a powerful vision-language model. Despite CLIP's proficiency, it suffers from view-dependent predictions and inherent bias, limiting its effectiveness. We propose a novel method that addresses these issues by leveraging multiple views near target objects, guided by Class Activation Mapping (CAM) of the classifier, and debiasing pseudo-labels derived from CLIP predictions. Our Classifier-guided CLIP Distillation (CCD) enables selecting multiple local views without extra labels and debiasing predictions to enhance classification performance. Experimental results validate our method's superiority over existing techniques across diverse datasets. The code is available at https://github.com/k0u-id/CCD.

Diffusion models are prone to exactly reproduce images from the training data. This exact reproduction of the training data is concerning as it can lead to copyright infringement and/or leakage of privacy-sensitive information. In this paper, we present a novel perspective on the memorization phenomenon and propose a simple yet effective approach to mitigate it. We argue that memorization occurs because of an attraction basin in the denoising process which steers the diffusion trajectory towards a memorized image. However, this can be mitigated by guiding the diffusion trajectory away from the attraction basin by not applying classifier-free guidance until an ideal transition point occurs from which classifier-free guidance is applied. This leads to the generation of non-memorized images that are high in image quality and well-aligned with the conditioning mechanism. To further improve on this, we present a new guidance technique, opposite guidance, that escapes the attraction basin sooner in the denoising process. We demonstrate the existence of attraction basins in various scenarios in which memorization occurs, and we show that our proposed approach successfully mitigates memorization.

Denoising is a crucial step in many video processing pipelines such as in interactive editing, where high quality, speed, and user control are essential. While recent approaches achieve significant improvements in denoising quality by leveraging deep learning, they are prone to unexpected failures due to discrepancies between training data distributions and the wide variety of noise patterns found in real-world videos. These methods also tend to be slow and lack user control. In contrast, traditional denoising methods perform reliably on in-the-wild videos and run relatively quickly on modern hardware. However, they require manually tuning parameters for each input video, which is not only tedious but also requires skill. We bridge the gap between these two paradigms by proposing a differentiable denoising pipeline based on traditional methods. A neural network is then trained to predict the optimal denoising parameters for each specific input, resulting in a robust and efficient approach that also supports user control.

Navigating dynamic urban environments presents significant challenges for embodied agents, requiring advanced spatial reasoning and adherence to common-sense norms. Despite progress, existing visual navigation methods struggle in map-free or off-street settings, limiting the deployment of autonomous agents like last-mile delivery robots. To overcome these obstacles, we propose a scalable, data-driven approach for human-like urban navigation by training agents on thousands of hours of in-the-wild city walking and driving videos sourced from the web. We introduce a simple and scalable data processing pipeline that extracts action supervision from these videos, enabling large-scale imitation learning without costly annotations. Our model learns sophisticated navigation policies to handle diverse challenges and critical scenarios. Experimental results show that training on large-scale, diverse datasets significantly enhances navigation performance, surpassing current methods. This work shows the potential of using abundant online video data to develop robust navigation policies for embodied agents in dynamic urban settings.

Good datasets are essential for developing and benchmarking any machine learning system. Their importance is even more extreme for safety critical applications such as deepfake detection - the focus of this paper. Here we reveal that two of the most widely used audio-video deepfake datasets suffer from a previously unidentified spurious feature: the leading silence. Fake videos start with a very brief moment of silence and, on the basis of this feature alone, we can separate the real and fake samples almost perfectly. As such, previous audio-only and audio-video models exploit the presence of silence in the fake videos and consequently perform worse when the leading silence is removed. To circumvent latching on such an unwanted artifact and possibly other unrevealed ones, we propose a shift from supervised to unsupervised learning by training models exclusively on real data. We show that by aligning self-supervised audio-video representations we remove the risk of relying on dataset-specific biases and improve robustness in deepfake detection.

Tool tracking in surgical videos is essential for advancing computer-assisted interventions, such as skill assessment, safety zone estimation, and human-machine collaboration. However, the lack of context-rich datasets limits AI applications in this field. Existing datasets rely on overly generic tracking formalizations that fail to capture surgical-specific dynamics, such as tools moving out of the camera's view or exiting the body. This results in less clinically relevant trajectories and a lack of flexibility for real-world surgical applications. Methods trained on these datasets often struggle with visual challenges such as smoke, reflection, and bleeding, further exposing the limitations of current approaches. We introduce CholecTrack20, a specialized dataset for multi-class, multi-tool tracking in surgical procedures. It redefines tracking formalization with three perspectives: (1) intraoperative, (2) intracorporeal, and (3) visibility, enabling adaptable and clinically meaningful tool trajectories. The dataset comprises 20 full-length surgical videos, annotated at 1 fps, yielding over 35K frames and 65K labeled tool instances. Annotations include spatial location, category, identity, operator, phase, and visual challenges. Benchmarking state-of-the-art methods on CholecTrack20 reveals significant performance gaps, with current approaches (<45% HOTA) failing to meet the accuracy required for clinical translation. These findings motivate the need for advanced and intuitive tracking algorithms and establish CholecTrack20 as a foundation for developing robust AI-driven surgical assistance systems.

Correct use of electrical appliances has significantly improved human life quality. Unlike simple tools that can be manipulated with common sense, different parts of electrical appliances have specific functions defined by manufacturers. If we want the robot to heat bread by microwave, we should enable them to review the microwave's manual first. From the manual, it can learn about component functions, interaction methods, and representative task steps about appliances. However, previous manual-related works remain limited to question-answering tasks while existing manipulation researchers ignore the manual's important role and fail to comprehend multi-page manuals. In this paper, we propose the first manual-based appliance manipulation benchmark CheckManual. Specifically, we design a large model-assisted human-revised data generation pipeline to create manuals based on CAD appliance models. With these manuals, we establish novel manual-based manipulation challenges, metrics, and simulator environments for model performance evaluation. Furthermore, we propose the first manual-based manipulation planning model ManualPlan to set up a group of baselines for the CheckManual benchmark.

Despite recent advances in deep texture recognition, existing methods still lack representational diversity and struggle to capture and preserve discriminative cues across stages of representation hierarchies. Moreover, many rely on loss formulations that prioritize recognition accuracy while overlooking spatial coherence and statistical consistency in the feature space. To address these issues, we propose three key innovations: Stochastic Local Texture Masking (SLTM), a regularization strategy that randomly occludes small texture patches to promote the learning of broader spatial and contextual dependencies; the Chebyshev Attention Depth Permutation Texture Network (CAPTN), a novel architecture that learns expressive and persistent Latent Texture Attribute (LTA) representations. CAPTN integrates a Texture Frequency Attention (TFA) module that generates LTAs and enables frequency-aware interpretability, a Dual Depth Permutation (D2P) module to expose complementary channel adjacency patterns, and Learnable Chebyshev Polynomials (LCPs) to model high-order orderless LTA transformations via recursive Chebyshev basis expansion; and a Latent Texture Attribute Loss that jointly optimizes classification accuracy, statistical alignment, and spatial fidelity. CAPTN supports end-to-end training without relying on fine-tuned CNN backbones and achieves state-of-the-art performance on several texture and material recognition benchmarks. (Code: https://github.com/RavishankarEvani/CAPTN)

Person re-identification (ReID) is the task of matching individuals across different camera views. Existing approaches typically employ neural networks to extract discriminative features, ranking gallery images based on their similarities to probe images. While effective, these methods are often enhanced through re-ranking, a post-processing step that refines initial retrieval results without requiring additional model training. However, current re-ranking methods mostly rely on k-nearest neighbor search to extract similar images that might have the same identity as the query, which is time-consuming with a high computation burden, limiting their applications in reality. We rethink the effect of the k-nearest neighbor search and introduce the Chebyshev's Theorem-guided Graph Re-ranking (Cheb-GR) method, which adopts the adaptive neighbor search guided by Chebyshev's Theorem over the k-nearest neighbor search for efficient neighbor selection. Our method leverages graph convolution operations to refine image features and achieve robust re-ranking, leading to enhanced retrieval performance. Furthermore, we provide a theoretical analysis based on Chebyshev's Inequality to elucidate the factors contributing to the strong performance of the proposed method. Our method significantly reduces the computation costs while maintaining relatively strong performance. Through extensive experiments in both general and cross-domain settings, we demonstrate the effectiveness of Cheb-GR and its potential for real-world applications.

Mobile health (mHealth) has emerged as a transformative solution to enhance healthcare accessibility and affordability, particularly in resource-constrained regions and low-to-middle-income countries.mHealth leverages mobile platforms to improve healthcare accessibility, addressing radiologist shortages in low-resource settings by enabling remote diagnosis and consultation through mobile devices. Mobile phones allow healthcare workers to transmit radiographic images, such as chest X-rays (CXR), to specialists or AI-driven models for interpretation. However, AI-based diagnosis using CXR images shared via apps like WhatsApp suffers from reduced predictability and explainability due to compression artifacts, and there is a lack of datasets to systematically study these challenges. To address this, we introduce CheXwhatsApp, a dataset of 141,804 paired original and WhatsApp-compressed CXR images. We present a benchmarking study which shows the dataset improves prediction stability and explainability of state-of-the-art models by up to 80%, while also enhancing localization performance. CheXwhatsApp is open-sourced to support advancements in mHealth applications for CXR analysis.

Humans can develop internal world models that encode common sense knowledge, telling them how the world works and predicting the consequences of their actions. This concept has emerged as a promising direction for establishing general-purpose machine-learning models in recent preliminary works, e.g., for visual representation learning. In this paper, we present CheXWorld, the first effort towards a self-supervised world model for radiographic images. Specifically, our work develops a unified framework that simultaneously models three aspects of medical knowledge essential for qualified radiologists, including 1) local anatomical structures describing the fine-grained characteristics of local tissues (e.g., architectures, shapes, and textures); 2) global anatomical layouts describing the global organization of the human body (e.g., layouts of organs and skeletons); and 3) domain variations that encourage CheXWorld to model the transitions across different appearance domains of radiographs (e.g., varying clarity, contrast, and exposure caused by collecting radiographs from different hospitals, devices, or patients). Empirically, we design tailored qualitative and quantitative analyses, revealing that CheXWorld successfully captures these three dimensions of medical knowledge. Furthermore, transfer learning experiments across eight medical image classification and segmentation benchmarks showcase that CheXWorld significantly outperforms existing SSL methods and large-scale medical foundation models. Code & pre-trained models are available at https://github.com/LeapLabTHU/CheXWorld.

Numerous methods have been proposed to detect, estimate, and analyze properties of people in images, including the estimation of 3D pose, shape, contact, human-object interaction, emotion, and more. While widely applicable in vision and other areas, such methods require expert knowledge to select, use, and interpret the results. To address this, we introduce ChatHuman, a language-driven system that combines and integrates the skills of these specialized methods. ChatHuman functions as an assistant proficient in utilizing, analyzing, and interacting with tools specific to 3D human tasks, adeptly discussing and resolving related challenges. Built on a Large Language Model (LLM) framework, ChatHuman is trained to autonomously select, apply, and interpret a diverse set of tools in response to user inputs. Adapting LLMs to 3D human tasks presents challenges, including the need for domain-specific knowledge and the ability to interpret complex 3D outputs. The innovative features of ChatHuman include leveraging academic publications to instruct the LLM how to use the tools, employing a retrieval-augmented generation model to create in-context learning examples for managing new tools, and effectively discriminating and integrating tool results to enhance tasks related to 3D humans by transforming specialized 3D outputs into comprehensible formats. Our experiments demonstrate that ChatHuman surpasses existing models in both tool selection accuracy and overall performance across various 3D human tasks, and it supports interactive chatting with users. ChatHuman represents a significant step toward consolidating diverse analytical methods into a unified, robust system for 3D human tasks.

Despite the significant advancements in text-to-image (T2I) generative models, users often face a trial-and-error challenge in practical scenarios. This challenge arises from the complexity and uncertainty of tedious steps such as crafting suitable prompts, selecting appropriate models, and configuring specific arguments, making users resort to labor-intensive attempts for desired images. This paper proposes Automatic T2I generation, which aims to automate these tedious steps, allowing users to simply describe their needs in a freestyle chatting way. To systematically study this problem, we first introduce ChatGenBench, a novel benchmark designed for Automatic T2I. It features high-quality paired data with diverse freestyle inputs, enabling comprehensive evaluation of automatic T2I models across all steps. Additionally, recognizing Automatic T2I as a complex multi-step reasoning task, we propose ChatGen-Evo, a multi-stage evolution strategy that progressively equips models with essential automation skills. Through extensive evaluation across step-wise accuracy and image quality, ChatGen-Evo significantly enhances performance over various baselines. Our evaluation also uncovers valuable insights for advancing automatic T2I. All our data, code and models will be publicly available.

We introduce ChatGarment, a novel approach that leverages large vision-language models (VLMs) to automate the estimation, generation, and editing of 3D garment sewing patterns from images or text descriptions. Unlike previous methods that often lack robustness and interactive editing capabilities, ChatGarment finetunes a VLM to produce GarmentCode, a JSON-based, language-friendly format for 2D sewing patterns, enabling both estimating and editing from images and text instructions. To optimize performance, we refine GarmentCode by expanding its support for more diverse garment types and simplifying its structure, making it more efficient for VLM finetuning. Additionally, we develop an automated data construction pipeline to generate a large-scale dataset of image-to-sewing-pattern and text-to-sewing-pattern pairs, empowering ChatGarment with strong generalization across various garment types. Extensive evaluations demonstrate ChatGarment's ability to accurately reconstruct, generate, and edit garments from multimodal inputs, highlighting its potential to revolutionize workflows in fashion and gaming applications.

Scalable Vector Graphics (SVG) has become the de facto standard for vector graphics in digital design, offering resolution independence and precise control over individual elements. Despite their advantages, creating high-quality SVG content remains challenging, as it demands technical expertise with professional editing software and a considerable time investment to craft complex shapes. Recent text-to-SVG generation methods aim to make vector graphics creation more accessible, but they still encounter limitations in shape regularity, generalization ability, and expressiveness. To address these challenges, we introduce Chat2SVG, a hybrid framework that combines the strengths of Large Language Models (LLMs) and image diffusion models for text-to-SVG generation. Our approach first uses an LLM to generate semantically meaningful SVG templates from basic geometric primitives. Guided by image diffusion models, a dual-stage optimization pipeline refines paths in latent space and adjusts point coordinates to enhance geometric complexity. Extensive experiments show that Chat2SVG outperforms existing methods in visual fidelity, path regularity, and semantic alignment. Additionally, our system enables intuitive editing through natural language instructions, making professional vector graphics creation accessible to all users. Our code is available at https://chat2svg.github.io/.

Traditional text-based person retrieval (TPR) relies on a single-shot text as query to retrieve the target person, assuming that the query completely captures the user's search intent. However, in real-world scenarios, it can be challenging to ensure the information completeness of such single-shot text. To address this limitation, we propose chat-based person retrieval (ChatPR), a new paradigm that takes an interactive dialogue as query to perform the person retrieval, engaging the user in conversational context to progressively refine the query for accurate person retrieval. The primary challenge in ChatPR is the lack of available dialogue-image paired data. To overcome this challenge, we establish ChatPedes, the first dataset designed for ChatPR, which is constructed by leveraging large language models to automate the question generation and simulate user responses. Additionally, to bridge the modality gap between dialogues and images, we propose a dialogue-refined cross-modal alignment (DiaNA) framework, which leverages two adaptive attribute refiners to bottleneck the conversational and visual information for fine-grained cross-modal alignment. Moreover, we propose a dialogue-specific data augmentation strategy, random round retaining, to further enhance the model's generalization ability across varying dialogue lengths. Extensive experiments demonstrate that DiaNA significantly outperforms existing TPR approaches, highlighting the effectiveness of conversational interactions for person retrieval.