CVPR 2025 Papers

Visual Document Understanding has become essential with the increase of text-rich visual content. This field poses significant challenges due to the need for effective integration of visual perception and textual comprehension, particularly across diverse document types with complex layouts. Moreover, existing fine-tuning datasets for this domain often fall short in providing the detailed contextual information for robust understanding, leading to hallucinations and limited comprehension of spatial relationships among visual elements. To address these challenges, we propose an innovative pipeline that utilizes adaptive generation of markup languages, such as Markdown, JSON, HTML, and TiKZ, to build highly structured document representations and deliver contextually-grounded responses. We introduce two fine-grained structured datasets: DocMark-Pile, comprising approximately 3.8M pretraining data pairs for document parsing, and DocMark-Instruct, featuring 624k fine-tuning data annotations for grounded instruction following.Extensive experiments demonstrate that our proposed model significantly outperforms existing state-of-the-art MLLMs across a range of visual document understanding benchmarks, facilitating advanced reasoning and comprehension capabilities in complex visual scenarios.

Multimodal large language models (MLLMs) have enabled open-world visual understanding by injecting visual input as extra tokens into large language models (LLMs) as contexts. However, when the visual input changes from a single image to a long video, the above paradigm encounters difficulty because the vast amount of video tokens has significantly exceeded the maximal capacity of MLLMs. Therefore, existing video-based MLLMs are mostly established upon sampling a small portion of tokens from input data, which can cause key information to be lost and thus produce incorrect answers. This paper presents a simple yet effective algorithm named Adaptive Keyframe Sampling (AKS). It inserts a plug-and-play module known as keyframe selection, which aims to maximize the useful information with a fixed number of video tokens. We formulate keyframe selection as an optimization involving (1) the relevance between the keyframes and the prompt, and (2) the coverage of the keyframes over the video, and present an adaptive algorithm to approximate the best solution. Experiments on two long video understanding benchmarks validate that AKS improves video QA accuracy (beyond strong baselines) upon selecting informative keyframes. Our study reveals the importance of information pre-filtering in video-based MLLMs. Our code are available at https://github.com/ncTimTang/AKS

In the domain of audio-visual event perception, which focuses on the temporal localization and classification of events across distinct modalities (audio and visual), existing approaches are constrained by the vocabulary available in their training data. This limitation significantly impedes their capacity to generalize to novel, unseen event categories. Furthermore, the annotation process for this task is labor-intensive, requiring extensive manual labeling across modalities and temporal segments, limiting the scalability of current methods. Current state-of-the-art models ignore the shifts in event distributions over time, reducing their ability to adjust to changing video dynamics. Additionally, previous methods rely on late fusion to combine audio and visual information. While straightforward, this approach results in a significant loss of multimodal interactions. To address these challenges, we propose Audio-Visual Adaptive Video Analysis (AV2A), a model-agnostic approach that requires no further training and integrates a score-level fusion technique to retain richer multimodal interactions. AV2A also includes a within-video label shift algorithm, leveraging input video data and predictions from prior frames to dynamically adjust event distributions for subsequent frames. Moreover, we present the first training-free, open-vocabulary baseline for audio-visual event perception, demonstrating that AV2A achieves substantial improvements over naive training-free baselines. We demonstrate the effectiveness of AV2A on both zero-shot and weakly-supervised state-of-the-art methods, achieving notable improvements in performance metrics over existing approaches.

The ability to adapt to varying observation lengths is crucial for human trajectory prediction tasks, particularly in scenarios with limited observation lengths or missing data. Existing approaches mainly focus on introducing novel architectures or additional structural components, which substantially increase model complexity and present challenges for integration into existing models. We argue that current network architectures are sufficiently sophisticated to handle the Observation Length Shift problem, with the key challenge lying in improving feature representation for trajectories with limited lengths. To tackle this issue, we introduce a general and effective contrastive learning approach, called Contrastive Learning for Length Shift (CLLS). By incorporating contrastive learning during the training phase, our method encourages the model to extract length-invariant features, thus mitigating the impact of observation length variations. Furthermore, to better accommodate length adaptation tasks, we introduce a lightweight RNN network that, combined with CLLS, achieves state-of-the-art performance in both general prediction and observation length shift tasks. Experimental results demonstrate that our approach outperforms existing methods across multiple widely-used trajectory prediction datasets.

Diffusion-based Text-to-Image (T2I) models have demonstrated significant potential in image restoration. However, existing models continue to grapple with challenges such as complex training and prompt design. We introduce a new perspective for improving image restoration by injecting knowledge from pretrained vision-language models into current T2I models. We empirically show that the degradation and content representations in BLIP-2 can be linearly separated, providing promising degradation guidance for image restoration. Specifically, the Feature Difference Instruction (FDI) is first extracted by Q-Formers through a simple subtraction operation based on reference image pairs. Then, we propose a multi-scale FDI adapter to decouple the degradation style and corrupted artifacts, and inject the styleflow exclusively into specific blocks through adapter-tuning, thereby preventing noise interference and eschewing the need for cumbersome weight retraining. In this way, we can train various task-specific adapters according to different degradations, achieving rich detail enhancement in the restoration results. Furthermore, the proposed FDI adapters have attractive properties of practical value, such as composability and generalization ability for all-in-one and mixed-degradation restoration. Extensive experiments under various settings demonstrate that our method has promising repairing quality over 10 image restoration tasks and a wide range of other applications.

Point cloud video understanding is becoming increasingly important in fields such as robotics, autonomous driving, and augmented reality, as they can accurately represent object motion and environmental changes. Despite the progress made in self-supervised learning methods for point cloud video understanding, the limited availability of 4D data and the high computational cost of training 4D-specific models remain significant obstacles. In this paper, we investigate the potential of transferring pre-trained static 3D point cloud models to the 4D domain, identifying the limitations of static models that capture only spatial information while neglecting temporal dynamics. To address this, we propose a novel Cross-frame Spatio-temporal Adaptation (CSA) strategy by introducing the Point Tube Adapter as the embedding layer and the Geometric Constraint Temporal Adapter (GCTA) to enforce temporal consistency across frames. This strategy extracts both short-term and long-term temporal dynamics, effectively integrating them with spatial features and enriching the model's understanding of temporal changes in point cloud videos. Extensive experiments on 3D action and gesture recognition tasks demonstrate that our method achieves state-of-the-art performance, establishing its effectiveness for point cloud video understanding.

Current deep homography estimation methods are typically constrained to processing low-resolution image pairs due to network architecture and computational limitations. For high-resolution images, downsampling is often required, which can greatly degrade estimation accuracy. In contrast, image matching methods, which match pixels and compute homography from correspondences, provide greater resolution flexibility. So in this work, we revisit the traditional image matching paradigm for homography estimation and propose GFNet, a Grid Flow regression Network that adapts the high-accuracy dense matching framework for homography estimation while enhancing efficiency through a grid-based strategy--estimating flow only over a coarse grid by leveraging homography's global smoothness. We demonstrate the effectiveness of GFNet on a wide range of experiments on multiple datasets, including the common scene MSCOCO, multimodal datasets VIS-IR and GoogleMap, and the dynamic scene VIRAT. Notably, on 448x448 GoogleMap, GFNet achieves an improvement of +13.5% in auc@3 while reducing MACs by ~47% compared to the SOTA dense matching method. Additionally, it shows a 1.8ximprovement in auc@3 over the SOTA deep homography method. Code is available at \textcolor[rgb] 0.95, 0.08, 0.58 https://github.com/KN-Zhang/GFNet .

We propose Adapter Merging with Centroid Prototype Mapping (ACMap), an exemplar-free framework for class-incremental learning (CIL) that addresses both catastrophic forgetting and scalability. While existing methods trade-off between inference time and accuracy, ACMap consolidates task-specific adapters into a single adapter, ensuring constant inference time across tasks without compromising accuracy.The framework employs adapter merging to build a shared subspace that aligns task representations and mitigates forgetting, while centroid prototype mapping maintains high accuracy through consistent adaptation in the shared subspace. To further improve scalability, an early stopping strategy limits adapter merging as tasks increase. Extensive experiments on five benchmark datasets demonstrate that ACMap matches state-of-the-art accuracy while maintaining inference speed comparable to the fastest existing methods.

Most Multi-view Clustering approaches assume that all views are available for clustering. However, this assumption is often unrealistic as views are incrementally accumulated over time, leading to a need for continual multi-view clustering (CMVC) methods. Current approaches to CMVC leverage late fusion-based approaches, where a new model is typically learned individually for each view to obtain the corresponding partition matrix, and then used to update a consensus matrix via a moving average. These approaches are prone to view-specific noise and struggle to adapt to large gaps between different views. To address these shortcomings, we reconsider CMVC from the perspective of domain adaptation and propose AdaptCMVC, which learns how to incrementally accumulate knowledge of new views as they become available and prevents catastrophic forgetting. Specifically, a self-training framework is introduced to extend the model to new views, particularly designed to be robust to view-specific noise. Further, to combat catastrophic forgetting, a structure alignment mechanism is proposed to enable the model to explore the global group structure across multiple views. Experiments on several multi-view benchmarks demonstrate the effectiveness of our proposed method on the CMVC task. The code is available at: AdaptCMVC.

Recently, model merging methods have demonstrated powerful strengths in combining abilities on various tasks from multiple Large Language Models (LLMs). While previous model merging methods mainly focus on merging homogeneous models with identical architecture, they meet challenges when dealing with Multimodal Large Language Models (MLLMs) with inherent heterogeneous property, including differences in model architecture and the asymmetry in the parameter space. In this work, we propose AdaMMS, a novel model merging method tailored for heterogeneous MLLMs. Our method tackles the challenges in three steps: mapping, merging and searching. Specifically, we first design mapping function between models to apply model merging on MLLMs with different architecture. Then we apply linear interpolation on model weights to actively adapt the asymmetry in the heterogeneous MLLMs. Finally in the hyper-parameter searching step, we propose an unsupervised hyper-parameter selection method for model merging. As the first model merging method capable of merging heterogeneous MLLMs without labeled data, extensive experiments on various model combinations demonstrated that AdaMMS outperforms previous model merging methods on various vision-language benchmarks.

Adapting Multi-modal Large Language Models (MLLMs) to target tasks often suffers from catastrophic forgetting, where acquiring new task-specific knowledge compromises performance on pre-trained tasks. In this paper, we introduce AdaDARE-\gamma, an efficient approach that alleviates catastrophic forgetting by controllably injecting new task-specific knowledge through adaptive parameter selection from fine-tuned models without requiring retraining procedures. This approach consists two key innovations: (1) an adaptive parameter selection mechanism that identifies and retains the most task-relevant parameters from fine-tuned models, and (2) a controlled task-specific information injection strategy that precisely balances the preservation of pre-trained knowledge with the acquisition of new capabilities. Theoretical analysis proves the optimality of our parameter selection strategy and establishes bounds for the task-specific information injection factor. Extensive experiments on InstructBLIP and LLaVA-1.5 across image captioning and visual question answering tasks demonstrate that AdaDARE-\gamma establishes new state-of-the-art results in balancing model performance. Specifically, it maintains 98.2% of pre-training effectiveness on original tasks while achieving 98.7% of standard fine-tuning performance on target tasks.

Incorporating transformer models into edge devices poses a significant challenge due to the computational demands of adapting these large models across diverse applications. Parameter-efficient tuning (PET) methods (e.g. LoRA, Adapter, Visual Prompt Tuning, etc.) allow for targeted adaptation by modifying only small parts of the transformer model. However, adapting to dynamic unlabeled target distributions at the test time remains complex. To address this, we introduce AdMiT: Adaptive Multi -Source Tuning in Dynamic Environments. AdMiT innovates by pre-training a set of PET modules, each optimized for different source distributions or tasks, and dynamically selecting and integrating a sparse subset of relevant modules when encountering a new, few-shot, unlabeled target distribution. This integration leverages Kernel Mean Embedding (KME)-based matching to align the target distribution with relevant source knowledge efficiently, without requiring additional routing networks or hyperparameter tuning. AdMiT achieves adaptation with a single inference step, making it particularly suitable for resource-constrained edge deployments. Furthermore, AdMiT preserves privacy by performing an adaptation locally on each edge device, without the need for data exchange. Our theoretical analysis establishes guarantees for AdMiT's generalization, while extensive benchmarks demonstrate that AdMiT consistently outperforms other PET methods across a range of tasks, achieving robust and efficient adaptation.

We introduce ActiveGAMER, an active mapping system that utilizes 3D Gaussian Splatting (3DGS) to achieve high-quality, real-time scene mapping and exploration. Unlike traditional NeRF-based methods, which are computationally demanding and restrict active mapping performance, our approach leverages the efficient rendering capabilities of 3DGS, allowing effective and efficient exploration in complex environments.The core of our system is a rendering-based information gain module that dynamically identifies the most informative viewpoints for next-best-view planning, enhancing both geometric and photometric reconstruction accuracy. ActiveGAMER also integrates a carefully balanced framework, combining coarse-to-fine exploration, post-refinement, and a global-local keyframe selection strategy to maximize reconstruction completeness and fidelity.Our system autonomously explores and reconstructs environments with state-of-the-art geometric and photometric accuracy and completeness, significantly surpassing existing approaches in both aspects. Extensive evaluations on benchmark datasets such as Replica and MP3D highlight ActiveGAMER's effectiveness in active mapping tasks.

Conventional frame-based imaging for active stereo systems has encountered major challenges in fast-motion scenarios. However, how to design a novel paradigm for high-speed depth sensing still remains an open issue. In this paper, we propose a novel problem setting, namely active event-based stereo vision, which provides the first insight of integrating binocular event cameras and an infrared projector for high-speed depth sensing. Technically, we first build a stereo camera prototype system and present a real-world dataset with over 21.5k spatiotemporal synchronized labels at 15 Hz, while also creating a realistic synthetic dataset with stereo event streams and 23.8k synchronized labels at 20 Hz. Then, we propose ActiveEventNet, a lightweight yet effective active event-based stereo matching neural network that learns to generate high-quality dense disparity maps from stereo event streams with low latency. Experiments demonstrate that our ActiveEventNet outperforms state-of-the-art methods meanwhile significantly reducing computational complexity. Our solution offers superior depth sensing compared to conventional stereo cameras in high-speed scenes, while also achieving the inference speed of up to 150 FPS with our prototype. We believe that this novel paradigm will provide new insights into future depth sensing systems. Our project can be available at https://github.com/jianing-li/active_event_based_stereo.

Knowledge distillation (KD) is the de facto standard for compressing large-scale models into smaller ones. Prior works have explored ever more complex KD strategies involving different objective functions, teacher-ensembles, and weight inheritance. In this work we explore an alternative, yet simple approach---active data curation as effective distillation for contrastive multimodal pretraining. Our simple online batch selection method, ACID, outperforms strong KD baselines across various model-, data- and compute-configurations. Further, we find such an active data curation strategy to in fact be complementary to standard KD, and can be effectively combined to train highly performant inference-efficient models. Our simple and scalable pretraining framework, ACED, achieves state-of-the-art results across 27 zero-shot classification and retrieval tasks with upto 11% less inference FLOPs. We further demonstrate that our ACED models yield strong vision-encoders for training generative multimodal models in the LiT-Decoder setting, outperforming larger vision encoders for image-captioning and visual question-answering tasks.

This work tackles the challenge of 3D Class-Incremental Learning (CIL), where a model must learn to classify new 3D objects while retaining knowledge of previously learned classes. Existing methods often struggle with catastrophic forgetting, misclassifying old objects due to overreliance on shortcut local features. Our approach addresses this issue by learning a set of part concepts for part-aware features. Particularly, we only activate a small subset of part concepts for the feature representation of each part-aware feature. This facilitates better generalization across categories and mitigates catastrophic forgetting. We further improve the task-wise classification through a part relation-aware Transformer design. At last, we devise learnable affinities to fuse task-wise classification heads and avoid confusion among different tasks. We evaluate our method on three 3D CIL benchmarks, achieving state-of-the-art performance. Code is available at https://github.com/zhenyatian/ILPC.

Video action recognition involves interpreting both global context and specific details to accurately identify actions. While previous models are effective at capturing spatiotemporal features, they often lack a focused representation of key action details. To address this, we introduce \nameo, a framework designed for refining video action recognition through integrated global and local feature learning. Inspired by human visual cognition theory, our approach balances the focus on both broad contextual changes and action-specific details, minimizing the influence of irrelevant background noise. We first employ learnable action queries to selectively emphasize action-relevant regions without requiring region-specific labels. Next, these queries are learned by a local action streaming branch that enables progressive query propagation. Moreover, we introduce a parameter-free feature interaction mechanism for effective multi-scale interaction between global and local features with minimal additional overhead. Extensive experiments demonstrate that \name achieves state-of-the-art performance across multiple action recognition datasets, validating its effectiveness and robustness in handling action-relevant details.

Recently, pre-trained text-to-image (T2I) models have been extensively adopted for real-world image restoration because of their powerful generative prior. However, controlling these large models for image restoration usually requires a large number of high-quality images and immense computational resources for training, which is costly and not privacy-friendly. In this paper, we find that the well-trained large T2I model (i.e., Flux) is able to produce a variety of high-quality images aligned with real-world distributions, offering an unlimited supply of training samples to mitigate the above issue. Specifically, we proposed a training data construction pipeline for image restoration, namely FluxGen, which includes unconditional image generation, image selection, and degraded image simulation. A novel light-weighted adapter (FluxIR) with squeeze-and-excitation layers is also carefully designed to control the large Diffusion Transformer (DiT)-based T2I model so that reasonable details can be restored. Experiments demonstrate that our proposed method enables the Flux model to adapt effectively to real-world image restoration tasks, achieving superior scores and visual quality on both synthetic and real-world degradation datasets - at only about 8.5% of the training cost compared to current approaches.