CVPR 2025 论文和开源项目合集(Papers with Code)

CVPR 2025 decisions are now available on OpenReview！22.1% = 2878 / 13008

注1：欢迎各位大佬提交issue，分享CVPR 2025论文和开源项目！

注2：关于往年CV顶会论文以及其他优质CV论文和大盘点，详见： https://github.com/amusi/daily-paper-computer-vision

• ICCV 2025
• ECCV 2024
• CVPR 2024

欢迎扫码加入【CVer学术交流群】，可以获取CVPR 2025等最前沿工作！这是最大的计算机视觉AI知识星球！每日更新，第一时间分享最新最前沿的计算机视觉、AIGC、扩散模型、多模态、深度学习、自动驾驶、医疗影像和遥感等方向的学习资料，快加入学起来！

【CVPR 2025 论文开源目录】

• 3DGS(Gaussian Splatting)
• Agent)
• Avatars
• Backbone
• CLIPEVOS
• Mamba
• Embodied AI
• GAN
• GNN
• 多模态大语言模型(MLLM)
• 大语言模型(LLM)
• NAS
• OCR
• NeRF
• DETR
• 扩散模型(Diffusion Models)
• ReID(重识别)
• 长尾分布(Long-Tail)
• Vision Transformer
• 视觉和语言(Vision-Language)
• 自监督学习(Self-supervised Learning)
• 数据增强(Data Augmentation)
• 目标检测(Object Detection)
• 异常检测(Anomaly Detection)
• 目标跟踪(Visual Tracking)
• 语义分割(Semantic Segmentation)
• 实例分割(Instance Segmentation)
• 全景分割(Panoptic Segmentation)
• 医学图像(Medical Image)
• 医学图像分割(Medical Image Segmentation)
• 视频目标分割(Video Object Segmentation)
• 视频实例分割(Video Instance Segmentation)
• 参考图像分割(Referring Image Segmentation)
• 图像抠图(Image Matting)
• 图像编辑(Image Editing)
• Low-level Vision
• 超分辨率(Super-Resolution)
• 去噪(Denoising)
• 去模糊(Deblur)
• 自动驾驶(Autonomous Driving)
• 3D点云(3D Point Cloud)
• 3D目标检测(3D Object Detection)
• 3D语义分割(3D Semantic Segmentation)
• 3D目标跟踪(3D Object Tracking)
• 3D语义场景补全(3D Semantic Scene Completion)
• 3D配准(3D Registration)
• 3D人体姿态估计(3D Human Pose Estimation)
• 3D人体Mesh估计(3D Human Mesh Estimation)
• 3D Visual Grounding(3D视觉定位)
• 医学图像(Medical Image)
• 图像生成(Image Generation)
• 视频生成(Video Generation)
• 3D生成(3D Generation)
• 视频理解(Video Understanding)
• 行为检测(Action Detection)
• 具身智能(Embodied AI)
• 文本检测(Text Detection)
• 知识蒸馏(Knowledge Distillation)
• 模型剪枝(Model Pruning)
• 图像压缩(Image Compression)
• 三维重建(3D Reconstruction)
• 深度估计(Depth Estimation)
• 轨迹预测(Trajectory Prediction)
• 车道线检测(Lane Detection)
• 图像描述(Image Captioning)
• 视觉问答(Visual Question Answering)
• 手语识别(Sign Language Recognition)
• 视频预测(Video Prediction)
• 新视点合成(Novel View Synthesis)
• Zero-Shot Learning(零样本学习)
• 立体匹配(Stereo Matching)
• 特征匹配(Feature Matching)
• 暗光图像增强(Low-light Image Enhancement)
• 场景图生成(Scene Graph Generation)
• 风格迁移(Style Transfer)
• 隐式神经表示(Implicit Neural Representations)
• 图像质量评价(Image Quality Assessment)
• 视频质量评价(Video Quality Assessment)
• 压缩感知(Compressive Sensing)
• 数据集(Datasets)
• 新任务(New Tasks)
• 其他(Others)

3DGS(Gaussian Splatting)

Agent

SpiritSight Agent: Advanced GUI Agent with One Look

• Paper: https://arxiv.org/abs/2503.03196
• Code: https://hzhiyuan.github.io/SpiritSight-Agent

Avatars

Backbone

Building Vision Models upon Heat Conduction

• Paper: https://arxiv.org/abs/2405.16555
• Code: https://github.com/MzeroMiko/vHeat

LSNet: See Large, Focus Small

• Paper: https://arxiv.org/abs/2503.23135
• Code: https://github.com/jameslahm/lsnet

CLIP

Mamba

MambaVision: A Hybrid Mamba-Transformer Vision Backbone

• Paper: https://arxiv.org/abs/2407.08083
• Code: https://github.com/NVlabs/MambaVision

MobileMamba: Lightweight Multi-Receptive Visual Mamba Network

• Paper: https://arxiv.org/abs/2411.15941
• Code: https://github.com/lewandofskee/MobileMamba

MambaIC: State Space Models for High-Performance Learned Image Compression

• Paper: https://arxiv.org/abs/2503.12461
• Code: https://arxiv.org/abs/2503.12461

Embodied AI

CityWalker: Learning Embodied Urban Navigation from Web-Scale Videos

• Project: https://ai4ce.github.io/CityWalker/
• Paper: https://arxiv.org/abs/2411.17820
• Code: https://github.com/ai4ce/CityWalker

GAN

OCR

NeRF

DETR

Mr. DETR: Instructive Multi-Route Training for Detection Transformers

• Paper: https://arxiv.org/abs/2412.10028
• Code: https://github.com/Visual-AI/Mr.DETR

Prompt

多模态大语言模型(MLLM)

LSceneLLM: Enhancing Large 3D Scene Understanding Using Adaptive Visual Preferences

• Paper： https://arxiv.org/abs/2412.01292
• Code: https://github.com/Hoyyyaard/LSceneLLM

DynRefer: Delving into Region-level Multimodal Tasks via Dynamic Resolution

• Paper: https://arxiv.org/abs/2405.16071
• Code: https://github.com/callsys/DynRefer

Retrieval-Augmented Personalization for Multimodal Large Language Models

• Project Page: https://hoar012.github.io/RAP-Project/
• Paper: https://arxiv.org/abs/2410.13360
• Code: https://github.com/Hoar012/RAP-MLLM

BiomedCoOp: Learning to Prompt for Biomedical Vision-Language Models

• Paper: https://arxiv.org/abs/2411.15232
• Code: https://github.com/HealthX-Lab/BiomedCoOp

FlashSloth: Lightning Multimodal Large Language Models via Embedded Visual Compression

• Paper: https://arxiv.org/abs/2412.04317
• Code: https://github.com/codefanw/FlashSloth

MMRL: Multi-Modal Representation Learning for Vision-Language Models

• Paper: https://arxiv.org/abs/2503.08497
• Code: https://github.com/yunncheng/MMRL

PAVE: Patching and Adapting Video Large Language Models

• Paper: https://arxiv.org/abs/2503.19794
• Code: https://github.com/dragonlzm/PAVE

AdaMMS: Model Merging for Heterogeneous Multimodal Large Language Models with Unsupervised Coefficient Optimization

• Paper: https://arxiv.org/abs/2503.23733
• Code: https://github.com/THUNLP-MT/AdaMMS

大语言模型(LLM)

NAS

ReID(重识别)

From Poses to Identity: Training-Free Person Re-Identification via Feature Centralization

• Paper: https://arxiv.org/abs/2503.00938
• Code: https://github.com/yuanc3/Pose2ID

AirRoom: Objects Matter in Room Reidentification

• Project: https://sairlab.org/airroom/
• Paper: https://arxiv.org/abs/2503.01130

IDEA: Inverted Text with Cooperative Deformable Aggregation for Multi-modal Object Re-Identification

• Paper: https://arxiv.org/abs/2503.10324
• Code: https://github.com/924973292/IDEA

扩散模型(Diffusion Models)

TinyFusion: Diffusion Transformers Learned Shallow

• Paper: https://arxiv.org/abs/2412.01199
• Code: https://github.com/VainF/TinyFusion

DKDM: Data-Free Knowledge Distillation for Diffusion Models with Any Architecture

• Paper: https://arxiv.org/abs/2409.03550
• Code: https://github.com/qianlong0502/DKDM

Tiled Diffusion

• Homepage: https://madaror.github.io/tiled-diffusion.github.io/
• Paper: https://arxiv.org/abs/2412.15185
• Code: https://github.com/madaror/tiled-diffusion

Vision Transformer

视觉和语言(Vision-Language)

NLPrompt: Noise-Label Prompt Learning for Vision-Language Models

• Paper: https://arxiv.org/abs/2412.01256
• Code: https://github.com/qunovo/NLPrompt

PhysVLM: Enabling Visual Language Models to Understand Robotic Physical Reachability

• Paper: https://arxiv.org/abs/2503.08481
• Code: https://github.com/unira-zwj/PhysVLM

MMRL: Multi-Modal Representation Learning for Vision-Language Models

• Paper: https://arxiv.org/abs/2503.08497
• Code: https://github.com/yunncheng/MMRL

目标检测(Object Detection)

LLMDet: Learning Strong Open-Vocabulary Object Detectors under the Supervision of Large Language Models

• Paper: https://arxiv.org/abs/2501.18954
• Code：https://github.com/iSEE-Laboratory/LLMDet

Mr. DETR: Instructive Multi-Route Training for Detection Transformers

• Paper: https://arxiv.org/abs/2412.10028
• Code: https://github.com/Visual-AI/Mr.DETR

异常检测(Anomaly Detection)

目标跟踪(Object Tracking)

Multiple Object Tracking as ID Prediction

• Paper：https://arxiv.org/abs/2403.16848
• Code: https://github.com/MCG-NJU/MOTIP

Omnidirectional Multi-Object Tracking

• Paper:https://arxiv.org/abs/2503.04565
• Code:https://github.com/xifen523/OmniTrack

医学图像(Medical Image)

BrainMVP: Multi-modal Vision Pre-training for Medical Image Analysis

• Paper: https://arxiv.org/abs/2410.10604
• Code: https://github.com/shaohao011/BrainMVP

医学图像分割(Medical Image Segmentation)

Test-Time Domain Generalization via Universe Learning: A Multi-Graph Matching Approach for Medical Image Segmentation

• Paper: https://arxiv.org/abs/2503.13012
• Code: https://github.com/Yore0/TTDG-MGM

自动驾驶(Autonomous Driving)

LiMoE: Mixture of LiDAR Representation Learners from Automotive Scenes

• Project: https://ldkong.com/LiMoE
• Paper: https://arxiv.org/abs/2501.04004
• Code: https://github.com/Xiangxu-0103/LiMoE

3D点云(3D-Point-Cloud)

Unlocking Generalization Power in LiDAR Point Cloud Registration

• Paper: https://arxiv.org/abs/2503.10149
• Code: https://github.com/peakpang/UGP

3D目标检测(3D Object Detection)

3D语义分割(3D Semantic Segmentation)

Low-level Vision

超分辨率(Super-Resolution)

AESOP: Auto-Encoded Supervision for Perceptual Image Super-Resolution

• Paper: https://arxiv.org/abs/2412.00124
• Code: https://github.com/2minkyulee/AESOP-Auto-Encoded-Supervision-for-Perceptual-Image-Super-Resolution

去噪(Denoising)

图像去噪(Image Denoising)

3D人体姿态估计(3D Human Pose Estimation)

Reconstructing Humans with a Biomechanically Accurate Skeleton

• Homepage: https://isshikihugh.github.io/HSMR/
• Code: https://github.com/IsshikiHugh/HSMR

#3D Visual Grounding(3D视觉定位)

ProxyTransformation: Preshaping Point Cloud Manifold With Proxy Attention For 3D Visual Grounding

• Homepage: https://pqh22.github.io/projects/ProxyTransformation/index.html

• Code: https://github.com/pqh22/ProxyTransformation

• Paper: https://arxiv.org/abs/2502.19247

图像生成(Image Generation)

Reconstruction vs. Generation: Taming Optimization Dilemma in Latent Diffusion Models

• Paper: https://arxiv.org/abs/2501.01423
• Code: https://github.com/hustvl/LightningDiT

SleeperMark: Towards Robust Watermark against Fine-Tuning Text-to-image Diffusion Models

• Paper: https://arxiv.org/abs/2412.04852
• Code: https://github.com/taco-group/SleeperMark

TokenFlow: Unified Image Tokenizer for Multimodal Understanding and Generation

• Homepage: https://byteflow-ai.github.io/TokenFlow/
• Code: https://github.com/ByteFlow-AI/TokenFlow
• Paper:https://arxiv.org/abs/2412.03069

PAR: Parallelized Autoregressive Visual Generation

• Project: https://epiphqny.github.io/PAR-project/
• Paper: https://arxiv.org/abs/2412.15119
• Code: https://github.com/Epiphqny/PAR

Generative Photography: Scene-Consistent Camera Control for Realistic Text-to-Image Synthesis

• Project: https://generative-photography.github.io/project/
• Paper: https://arxiv.org/abs/2412.02168
• Code: https://github.com/pandayuanyu/generative-photography

OpenING: A Comprehensive Benchmark for Judging Open-ended Interleaved Image-Text Generation

• Project Page: https://opening-benchmark.github.io/
• Paper: https://arxiv.org/abs/2411.18499).
• Code: https://github.com/LanceZPF/OpenING

视频生成(Video Generation)

Identity-Preserving Text-to-Video Generation by Frequency Decomposition

• Paper: https://arxiv.org/abs/2411.17440
• Code: https://github.com/PKU-YuanGroup/ConsisID

Cinemo: Consistent and Controllable Image Animation with Motion Diffusion Models

• Paper: https://arxiv.org/abs/2407.15642
• Code: https://github.com/maxin-cn/Cinemo

X-Dyna: Expressive Dynamic Human Image Animation

• Paper: https://arxiv.org/abs/2501.10021
• Code: https://github.com/bytedance/X-Dyna

PhyT2V: LLM-Guided Iterative Self-Refinement for Physics-Grounded Text-to-Video Generation

• Paper: https://arxiv.org/pdf/2412.00596
• Code: https://github.com/pittisl/PhyT2V

Timestep Embedding Tells: It's Time to Cache for Video Diffusion Model

• Project: https://liewfeng.github.io/TeaCache/
• Paper: https://arxiv.org/abs/2411.19108
• Code: https://github.com/ali-vilab/TeaCache

AR-Diffusion: Asynchronous Video Generation with Auto-Regressive Diffusion

• Project: https://iva-mzsun.github.io/AR-Diffusion
• Paper: https://arxiv.org/abs/2503.07418
• Code: https://github.com/iva-mzsun/AR-Diffusion

图像编辑(Image Editing)

Edit Away and My Face Will not Stay: Personal Biometric Defense against Malicious Generative Editing

• Paper: https://arxiv.org/abs/2411.16832
• Code: https://github.com/taco-group/FaceLock

h-Edit: Effective and Flexible Diffusion-Based Editing via Doob’s h-Transform

• Paper: https://arxiv.org/abs/2503.02187
• Code: https://github.com/nktoan/h-edit

视频编辑(Video Editing)

3D生成(3D Generation)

Generative Gaussian Splatting for Unbounded 3D City Generation

• Project: https://haozhexie.com/project/gaussian-city
• Paper: https://arxiv.org/abs/2406.06526
• Code: https://github.com/hzxie/GaussianCity

StdGEN: Semantic-Decomposed 3D Character Generation from Single Images

• Project: https://stdgen.github.io/
• Paper: https://arxiv.org/abs/2411.05738
• Code: https://github.com/hyz317/StdGEN

3D重建(3D Reconstruction)

Fast3R: Towards 3D Reconstruction of 1000+ Images in One Forward Pass

• Project: https://fast3r-3d.github.io/
• Paper: https://arxiv.org/abs/2501.13928

人体运动生成(Human Motion Generation)

SemGeoMo: Dynamic Contextual Human Motion Generation with Semantic and Geometric Guidance

• Project: https://4dvlab.github.io/project_page/semgeomo/
• Paper: https://arxiv.org/abs/2503.01291
• https://github.com/4DVLab/SemGeoMo

视频理解(Video Understanding)

Temporal Grounding Videos like Flipping Manga

• Paper: https://arxiv.org/abs/2411.10332
• Code: https://github.com/yongliang-wu/NumPro

具身智能(Embodied AI)

Universal Actions for Enhanced Embodied Foundation Models

• Project: https://2toinf.github.io/UniAct/
• Paper: https://arxiv.org/abs/2501.10105
• Code: https://github.com/2toinf/UniAct

PhysVLM: Enabling Visual Language Models to Understand Robotic Physical Reachability

• Paper: https://arxiv.org/abs/2503.08481
• Code: https://github.com/unira-zwj/PhysVLM

知识蒸馏(Knowledge Distillation)

深度估计(Depth Estimation)

DepthCrafter: Generating Consistent Long Depth Sequences for Open-world Videos

• Project: https://depthcrafter.github.io
• Paper: https://arxiv.org/abs/2409.02095
• Code: https://github.com/Tencent/DepthCrafter

MonSter: Marry Monodepth to Stereo Unleashes Power

• Paper: https://arxiv.org/abs/2501.08643
• Code: https://github.com/Junda24/MonSter

DEFOM-Stereo: Depth Foundation Model Based Stereo Matching

• Project: https://insta360-research-team.github.io/DEFOM-Stereo/
• Paper: https://arxiv.org/abs/2501.09466
• Code: https://github.com/Insta360-Research-Team/DEFOM-Stereo

立体匹配(Stereo Matching)

MonSter: Marry Monodepth to Stereo Unleashes Power

• Paper: https://arxiv.org/abs/2501.08643
• Code: https://github.com/Junda24/MonSter

暗光图像增强(Low-light Image Enhancement)

HVI: A New color space for Low-light Image Enhancement

• Paper: https://arxiv.org/abs/2502.20272
• Code: https://github.com/Fediory/HVI-CIDNet
• Demo: https://huggingface.co/spaces/Fediory/HVI-CIDNet_Low-light-Image-Enhancement_

ReDDiT: Efficient Diffusion as Low Light Enhancer

• Paper: https://arxiv.org/abs/2410.12346
• Code: https://github.com/lgz-0713/ReDDiT

图像压缩(Image Compression)](#IC)

MambaIC: State Space Models for High-Performance Learned Image Compression

• Paper: https://arxiv.org/abs/2503.12461
• Code: https://arxiv.org/abs/2503.12461

场景图生成(Scene Graph Generation)

风格迁移(Style Transfer)

StyleStudio: Text-Driven Style Transfer with Selective Control of Style Elements

• Project: https://stylestudio-official.github.io/
• Paper: https://arxiv.org/abs/2412.08503
• Code: https://github.com/Westlake-AGI-Lab/StyleStudio

图像质量评价(Image Quality Assessment)

Auto Cherry-Picker: Learning from High-quality Generative Data Driven by Language

• Homepage: https://yichengchen24.github.io/projects/autocherrypicker
• Paper: https://arxiv.org/pdf/2406.20085
• Code: https://github.com/yichengchen24/ACP

视频质量评价(Video Quality Assessment)

压缩感知(Compressive Sensing)

Using Powerful Prior Knowledge of Diffusion Model in Deep Unfolding Networks for Image Compressive Sensing

• Paper: https://arxiv.org/abs/2503.08429
• Code: https://github.com/FengodChen/DMP-DUN-CVPR2025

数据集(Datasets)

Objaverse++: Curated 3D Object Dataset with Quality Annotations

• Paper: https://arxiv.org/abs/2504.07334
• Code: https://github.com/TCXX/ObjaversePlusPlus

其他(Others)

DTGBrepGen: A Novel B-rep Generative Model through Decoupling Topology and Geometry

• Paper: https://arxiv.org/abs/2503.13110
• Code: https://github.com/jinli99/DTGBrepGen

Analyzing the Synthetic-to-Real Domain Gap in 3D Hand Pose Estimation

• Paper: https://arxiv.org/abs/2503.19307
• Code: https://github.com/delaprada/HandSynthesis.git

EVOS: Efficient Implicit Neural Training via EVOlutionary Selector

• Homepage: https://weixiang-zhang.github.io/proj-evos/
• Paper: https://arxiv.org/abs/2412.10153
• Code: https://github.com/zwx-open/EVOS-INR