CVPR2023已经放榜，今年有2360篇，接收率为25.78%。在CVPR2023正式会议召开前，为了让大家更快地获取和学习到计算机视觉前沿技术，极市对CVPR023 最新论文进行追踪，包括分研究方向的论文、代码汇总以及论文技术直播分享。

CVPR 2023 论文分方向整理目前在极市社区持续更新中，已累计更新了 919 篇，项目地址： https://www. cvmart.net/community/de tail/7422

以下是最近更新的 CVPR 2023 论文，涵盖神经网络结构、医学影像、ReId、图像去雾、异常检测等方向。

下载地址： https://www. cvmart.net/community/de tail/7520

2D目标检测(2D Object Detection)

[1]Mapping Degeneration Meets Label Evolution: Learning Infrared Small Target Detection with Single Point Supervision
paper： https:// arxiv.org/abs/2304.0148 4
code： https:// github.com/xinyiying/le sps

[2]Multi-view Adversarial Discriminator: Mine the Non-causal Factors for Object Detection in Unseen Domains
paper： https:// arxiv.org/abs/2304.0295 0

[3]Continual Detection Transformer for Incremental Object Detection
paper： https:// arxiv.org/abs/2304.0311 0

[4]DetCLIPv2: Scalable Open-Vocabulary Object Detection Pre-training via Word-Region Alignment
paper： https:// arxiv.org/abs/2304.0451 4

[5]Benchmarking the Physical-world Adversarial Robustness of Vehicle Detection
paper： https:// arxiv.org/abs/2304.0509 8

3D目标检测(3D object detection)

[1]Hierarchical Supervision and Shuffle Data Augmentation for 3D Semi-Supervised Object Detection
paper： https:// arxiv.org/abs/2304.0146 4
code： https:// github.com/azhuantou/hs sda

[2]Curricular Object Manipulation in LiDAR-based Object Detection
paper： https:// arxiv.org/abs/2304.0424 8
code： https:// github.com/zzy816/com

人物交互检测(HOI Detection)

[1]Instant-NVR: Instant Neural Volumetric Rendering for Human-object Interactions from Monocular RGBD Stream
paper： https:// arxiv.org/abs/2304.0318 4

[2]Relational Context Learning for Human-Object Interaction Detection
paper： https:// arxiv.org/abs/2304.0499 7

异常检测(Anomaly Detection)

[1]Robust Outlier Rejection for 3D Registration with Variational Bayes
paper： https:// arxiv.org/abs/2304.0151 4
code： https:// github.com/jiang-hb/vbr eg

[2]Video Event Restoration Based on Keyframes for Video Anomaly Detection
paper： https:// arxiv.org/abs/2304.0511 2

语义分割(Semantic Segmentation)

[1]DiGA: Distil to Generalize and then Adapt for Domain Adaptive Semantic Segmentation
paper： https:// arxiv.org/abs/2304.0222 2
code： https:// github.com/fy-vision/di ga

[2]Exploiting the Complementarity of 2D and 3D Networks to Address Domain-Shift in 3D Semantic Segmentation
paper： https:// arxiv.org/abs/2304.0299 1
code： https:// github.com/cvlab-unibo/ mm2d3d

[3]Federated Incremental Semantic Segmentation
paper： https:// arxiv.org/abs/2304.0462 0
code： https:// github.com/jiahuadong/f iss

[4]Continual Semantic Segmentation with Automatic Memory Sample Selection
paper： https:// arxiv.org/abs/2304.0501 5

深度估计(Depth Estimation)

[1]EGA-Depth: Efficient Guided Attention for Self-Supervised Multi-Camera Depth Estimation
paper： https:// arxiv.org/abs/2304.0336 9

[2]DualRefine: Self-Supervised Depth and Pose Estimation Through Iterative Epipolar Sampling and Refinement Toward Equilibrium
paper： https:// arxiv.org/abs/2304.0356 0
code： https:// github.com/antabangun/d ualrefine

人体解析/人体姿态估计(Human Parsing/Human Pose Estimation)

[1]A2J-Transformer: Anchor-to-Joint Transformer Network for 3D Interacting Hand Pose Estimation from a Single RGB Image
paper： https:// arxiv.org/abs/2304.0363 5
code： https:// github.com/changlongjia nggit/a2j-transformer

[2]Monocular 3D Human Pose Estimation for Sports Broadcasts using Partial Sports Field Registration
paper： https:// arxiv.org/abs/2304.0443 7
code： https:// github.com/tobibaum/par tialsportsfieldreg_3dhpe

[3]DeFeeNet: Consecutive 3D Human Motion Prediction with Deviation Feedback
paper： https:// arxiv.org/abs/2304.0449 6

视频处理(Video Processing)

[1]BiFormer: Learning Bilateral Motion Estimation via Bilateral Transformer for 4K Video Frame Interpolation
paper： https:// arxiv.org/abs/2304.0222 5
code： https:// github.com/junheum/bifo rmer

超分辨率(Super Resolution)

[1]Better "CMOS" Produces Clearer Images: Learning Space-Variant Blur Estimation for Blind Image Super-Resolution
paper： https:// arxiv.org/abs/2304.0354 2

图像复原/图像增强/图像重建(Image Restoration/Image Reconstruction)

[1]Generative Diffusion Prior for Unified Image Restoration and Enhancement
paper： https:// arxiv.org/abs/2304.0124 7

[2]CherryPicker: Semantic Skeletonization and Topological Reconstruction of Cherry Trees
paper： https:// arxiv.org/abs/2304.0470 8

图像去噪/去模糊/去雨去雾(Image Denoising)

[1]HyperCUT: Video Sequence from a Single Blurry Image using Unsupervised Ordering
paper： https:// arxiv.org/abs/2304.0168 6

[2]RIDCP: Revitalizing Real Image Dehazing via High-Quality
codebook Priors
paper： https:// arxiv.org/abs/2304.0399 4
code： https:// github.com/RQ-Wu/RIDCP_ dehazing

人脸识别/检测(Facial Recognition/Detection)

[1]Gradient Attention Balance Network: Mitigating Face Recognition Racial Bias via Gradient Attention
paper： https:// arxiv.org/abs/2304.0228 4

[2]Micron-BERT: BERT-based Facial Micro-Expression Recognition
paper： https:// arxiv.org/abs/2304.0319 5
code： https:// github.com/uark-cviu/mi cron-bert

人脸生成/合成/重建/编辑(Face Generation/Face Synthesis/Face Reconstruction/Face Editing)

[1]Learning Personalized High Quality Volumetric Head Avatars from Monocular RGB Videos
paper： https:// arxiv.org/abs/2304.0143 6

[2]StyleGAN Salon: Multi-View Latent Optimization for Pose-Invariant Hairstyle Transfer
paper： https:// arxiv.org/abs/2304.0274 4

[3]GANHead: Towards Generative Animatable Neural Head Avatars
paper： https:// arxiv.org/abs/2304.0395 0

目标跟踪(Object Tracking)

[1]Trace and Pace: Controllable Pedestrian Animation via Guided Trajectory Diffusion
paper： https:// arxiv.org/abs/2304.0189 3

[2]Unsupervised Sampling Promoting for Stochastic Human Trajectory Prediction
paper： https:// arxiv.org/abs/2304.0429 8
code： https:// github.com/viewsetting/ unsupervised_sampling_promoting

图像&视频检索/视频理解(Image&Video Retrieval/Video Understanding)

[1]Improving Image Recognition by Retrieving from Web-Scale Image-Text Data
paper： https:// arxiv.org/abs/2304.0517 3

行人重识别/检测(Re-Identification/Detection)

[1]PartMix: Regularization Strategy to Learn Part Discovery for Visible-Infrared Person Re-identification
paper： https:// arxiv.org/abs/2304.0153 7

[2]Shape-Erased Feature Learning for Visible-Infrared Person Re-Identification
paper： https:// arxiv.org/abs/2304.0420 5
code： https:// github.com/jiawei151/sg iel_vireid

图像/视频字幕(Image/Video Caption)

[1]Cross-Domain Image Captioning with Discriminative Finetuning
paper： https:// arxiv.org/abs/2304.0166 2
code： https:// github.com/facebookrese arch/EGG

[2]Model-Agnostic Gender Debiased Image Captioning
paper： https:// arxiv.org/abs/2304.0369 3

医学影像(Medical Imaging)

[1]Topology-Guided Multi-Class Cell Context Generation for Digital Pathology
paper： https:// arxiv.org/abs/2304.0225 5

[2]Deep Prototypical-Parts Ease Morphological Kidney Stone Identification and are Competitively Robust to Photometric Perturbations
paper： https:// arxiv.org/abs/2304.0407 7
code： https:// github.com/danielf29/pr ototipical_parts

[3]Coherent Concept-based Explanations in Medical Image and Its Application to Skin Lesion Diagnosis
paper： https:// arxiv.org/abs/2304.0457 9
code： https:// github.com/cristianopat ricio/coherent-cbe-skin

图像生成/图像合成(Image Generation/Image Synthesis)

[1]Toward Verifiable and Reproducible Human Evaluation for Text-to-Image Generation
paper： https:// arxiv.org/abs/2304.0181 6

[2]Few-shot Semantic Image Synthesis with Class Affinity Transfer
paper： https:// arxiv.org/abs/2304.0232 1

点云(Point Cloud)

[1]MEnsA: Mix-up Ensemble Average for Unsupervised Multi Target Domain Adaptation on 3D Point Clouds
paper： https:// arxiv.org/abs/2304.0155 4
code： https:// github.com/sinashish/me nsa_mtda

场景重建/视图合成/新视角合成(Novel View Synthesis)

[1]Lift3D: Synthesize 3D Training Data by Lifting 2D GAN to 3D Generative Radiance Field
paper： https:// arxiv.org/abs/2304.0352 6

[2]POEM: Reconstructing Hand in a Point Embedded Multi-view Stereo
paper： https:// arxiv.org/abs/2304.0403 8
code： https:// github.com/lixiny/poem

[3]Neural Residual Radiance Fields for Streamably Free-Viewpoint Videos
paper： https:// arxiv.org/abs/2304.0445 2

[4]Neural Lens Modeling
paper： https:// arxiv.org/abs/2304.0484 8

[5]One-Shot High-Fidelity Talking-Head Synthesis with Deformable Neural Radiance Field
paper： https:// arxiv.org/abs/2304.0509 7

[6]MonoHuman: Animatable Human Neural Field from Monocular Video
paper： https:// arxiv.org/abs/2304.0200 1

[7]GINA-3D: Learning to Generate Implicit Neural Assets in the Wild
paper： https:// arxiv.org/abs/2304.0216 3

[8]Neural Fields meet Explicit Geometric Representation for Inverse Rendering of Urban Scenes
paper： https:// arxiv.org/abs/2304.0326 6

文本检测/识别/理解(Text Detection/Recognition/Understanding)

[1]Towards Unified Scene Text Spotting based on Sequence Generation
paper： https:// arxiv.org/abs/2304.0343 5

神经网络结构设计(Neural Network Structure Design)

[1]SMPConv: Self-moving Point Representations for Continuous Convolution
paper： https:// arxiv.org/abs/2304.0233 0
code： https:// github.com/sangnekim/sm pconv

CNN

[1]VNE: An Effective Method for Improving Deep Representation by Manipulating Eigenvalue Distribution
paper： https:// arxiv.org/abs/2304.0143 4
code： https:// github.com/jaeill/CVPR2 3-VNE

Transformer

[1]METransformer: Radiology Report Generation by Transformer with Multiple Learnable Expert Tokens
paper： https:// arxiv.org/abs/2304.0221 1

[2]MethaneMapper: Spectral Absorption aware Hyperspectral Transformer for Methane Detection
paper： https:// arxiv.org/abs/2304.0276 7

[3]Visual Dependency Transformers: Dependency Tree Emerges from Reversed Attention
paper： https:// arxiv.org/abs/2304.0328 2
code： https:// github.com/dingmyu/depe ndencyvit

[4]Slide-Transformer: Hierarchical Vision Transformer with Local Self-Attention
paper： https:// arxiv.org/abs/2304.0423 7
code： https:// github.com/leaplabthu/s lide-transformer

图神经网络(GNN)

[1]Adversarially Robust Neural Architecture Search for Graph Neural Networks
paper： https:// arxiv.org/abs/2304.0416 8

归一化/正则化(Batch Normalization)

[1]Delving into Discrete Normalizing Flows on SO(3) Manifold for Probabilistic Rotation Modeling
paper： https:// arxiv.org/abs/2304.0393 7

模型训练/泛化(Model Training/Generalization)

[1]Re-thinking Model Inversion Attacks Against Deep Neural Networks
paper： https:// arxiv.org/abs/2304.0166 9

[2]Improved Test-Time Adaptation for Domain Generalization
paper： https:// arxiv.org/abs/2304.0449 4

长尾分布(Long-Tailed Distribution)

[1]Long-Tailed Visual Recognition via Self-Heterogeneous Integration with Knowledge Excavation
paper： https:// arxiv.org/abs/2304.0127 9
code： https:// github.com/jinyan-06/sh ike

视觉表征学习(Visual Representation Learning)

[1]HNeRV: A Hybrid Neural Representation for Videos
paper： https:// arxiv.org/abs/2304.0263 3
code： https:// github.com/haochen-rye/ hnerv

多模态学习(Multi-Modal Learning)

[1]Detecting and Grounding Multi-Modal Media Manipulation
paper： https:// arxiv.org/abs/2304.0255 6
code： https:// github.com/rshaojimmy/m ultimodal-deepfake

[2]Learning Instance-Level Representation for Large-Scale Multi-Modal Pretraining in E-commerce
paper： https:// arxiv.org/abs/2304.0285 3

[3]Vita-CLIP: Video and text adaptive CLIP via Multimodal Prompting
paper： https:// arxiv.org/abs/2304.0330 7
code： https:// github.com/talalwasim/v ita-clip

视觉-语言（Vision-language）

[1]Learning to Name Classes for Vision and Language Models
paper： https:// arxiv.org/abs/2304.0183 0

[2]VLPD: Context-Aware Pedestrian Detection via Vision-Language Semantic Self-Supervision
paper： https:// arxiv.org/abs/2304.0313 5
code： https:// github.com/lmy98129/vlp d

[3]CrowdCLIP: Unsupervised Crowd Counting via Vision-Language Model
paper： https:// arxiv.org/abs/2304.0423 1
code： https:// github.com/dk-liang/cro wdclip

[4]Improving Vision-and-Language Navigation by Generating Future-View Image Semantics
paper： https:// arxiv.org/abs/2304.0490 7

场景图生成(Scene Graph Generation)

[1]Devil's on the Edges: Selective Quad Attention for Scene Graph Generation
paper： https:// arxiv.org/abs/2304.0349 5

视觉推理/视觉问答(Visual Reasoning/VQA)

[1]Language Models are Causal Knowledge Extractors for Zero-shot Video Question Answering
paper： https:// arxiv.org/abs/2304.0375 4

数据集(Dataset)

[1]Uncurated Image-Text Datasets: Shedding Light on Demographic Bias
paper： https:// arxiv.org/abs/2304.0282 8
code： https:// github.com/noagarcia/ph ase

小样本学习/零样本学习(Few-shot Learning/Zero-shot Learning)

[1]Zero-shot Generative Model Adaptation via Image-specific Prompt Learning
paper： https:// arxiv.org/abs/2304.0311 9

迁移学习/domain/自适应(Transfer Learning/Domain Adaptation)

[1]DATE: Domain Adaptive Product Seeker for E-commerce
paper： https:// arxiv.org/abs/2304.0366 9

[2]Modernizing Old Photos Using Multiple References via Photorealistic Style Transfer
paper： https:// arxiv.org/abs/2304.0446 1

持续学习(Continual Learning/Life-long Learning)

[1]Asynchronous Federated Continual Learning
paper： https:// arxiv.org/abs/2304.0362 6
code： https:// github.com/lttm/fedspac e

[2]Exploring Data Geometry for Continual Learning
paper： https:// arxiv.org/abs/2304.0393 1

[3]Task Difficulty Aware Parameter Allocation & Regularization for Lifelong Learning
paper： https:// arxiv.org/abs/2304.0528 8
code： https:// github.com/wenjinw/par

[4]Online Distillation with Continual Learning for Cyclic Domain Shifts
paper： https:// arxiv.org/abs/2304.0123 9

视觉定位/位姿估计(Visual Localization/Pose Estimation)

[1]OrienterNet: Visual Localization in 2D Public Maps with Neural Matching
paper： https:// arxiv.org/abs/2304.0200 9

增量学习(Incremental Learning)

[1]On the Stability-Plasticity Dilemma of Class-Incremental Learning
paper： https:// arxiv.org/abs/2304.0166 3

[2]PCR: Proxy-based Contrastive Replay for Online Class-Incremental Continual Learning
paper： https:// arxiv.org/abs/2304.0440 8

强化学习(Reinforcement Learning)

[1]Reinforcement Learning-Based Black-Box Model Inversion Attacks
paper： https:// arxiv.org/abs/2304.0462 5

元学习(Meta Learning)

[1]Meta-causal Learning for Single Domain Generalization
paper： https:// arxiv.org/abs/2304.0370 9

[2]Meta Compositional Referring Expression Segmentation
paper： https:// arxiv.org/abs/2304.0441 5

[3]Meta-Learning with a Geometry-Adaptive Preconditioner
paper： https:// arxiv.org/abs/2304.0155 2
code： https:// github.com/suhyun777/cv pr23-gap

半监督学习/弱监督学习/无监督学习/自监督学习(Self-supervised Learning/Semi-supervised Learning)

[1]Weakly supervised segmentation with point annotations for histopathology images via contrast-based variational model
paper： https:// arxiv.org/abs/2304.0357 2

[2]Token Boosting for Robust Self-Supervised Visual Transformer Pre-training
paper： https:// arxiv.org/abs/2304.0417 5

[3]SOOD: Towards Semi-Supervised Oriented Object Detection
paper： https:// arxiv.org/abs/2304.0451 5
code： https:// github.com/hamperdredes /sood

[4]Defending Against Patch-based Backdoor Attacks on Self-Supervised Learning
paper： https:// arxiv.org/abs/2304.0148 2
code： https:// github.com/ucdvision/pa tchsearch

神经网络可解释性(Neural Network Interpretability)

[1]Gradient-based Uncertainty Attribution for Explainable Bayesian Deep Learning
paper： https:// arxiv.org/abs/2304.0482 4

图像计数(Image Counting)

[1]Density Map Distillation for Incremental Object Counting
paper： https:// arxiv.org/abs/2304.0525 5

其他

[1]Bridging the Gap between Model Explanations in Partially Annotated Multi-label Classification
paper： https:// arxiv.org/abs/2304.0180 4
code： https:// github.com/youngwk/brid gegapexplanationpamc

[2]Knowledge Combination to Learn Rotated Detection Without Rotated Annotation
paper： https:// arxiv.org/abs/2304.0219 9

[3]CloSET: Modeling Clothed Humans on Continuous Surface with Explicit Template Decomposition
paper： https:// arxiv.org/abs/2304.0316 7

[4]DC2: Dual-Camera Defocus Control by Learning to Refocus
paper： https:// arxiv.org/abs/2304.0328