CVPR2020在2月24日公布论文

CVPR2020在2月24日公布了所有接受论文ID，相关报道：1470篇！CVPR2020结果出炉，你中了吗？（附部分论文链接/开源代码/解读）。自论文ID公布以来，许多开发者都分享了自己的优秀工作。

从论文ID公布以来，极市一直在对CVPR进行实时跟进，本文是对CVPR2020论文整理和分类，均有论文链接，部分含开源代码，涵盖的方向有：目标检测、目标跟踪、图像分割、人脸识别、姿态估计、三维点云、视频分析、模型加速、GAN、OCR等方向。

为了方便大家阅读，小极已经将全部论文下载并打包。扫描下方二维码 关注 极市平台 公众号，回复 CVPR2020 即可获取下载链接。同时，可访问 极市社区，后续论文收录会在这里保持更新。关注极市平台，获取CVPR2020论文合集下载链接
声明：本文为极市平台原创整理，未经许可，不得擅自转载。

此外，我们也会在Github和极市社区上保持更新，欢迎大家关注：
https://github.com/extreme-assistant/cvpr2020/blob/master/CVPR2020.md
目录

1. 目标检测

2. 图像分割

3. 人脸识别

4. 目标跟踪

5. 三维点云/三维重建

6. 图像处理

7. 图像分类

8. 姿态估计/动作识别

9. 视频分析

10. OCR

11. GAN

12. 小样本/零样本

13. 弱监督/无监督/自监督

14. 行人跟踪/行人检测/ReID

15. 神经网络/模型加速/模型压缩

16. 超分辨率

17. 视觉常识/数据集/其他

 

目标检测

    Bridging the Gap Between Anchor-based and Anchor-free Detection via Adaptive Training Sample Selection

    论文地址：https://arxiv.org/abs/1912.02424

    代码：https://github.com/sfzhang15/ATSS

    Few-Shot Object Detection with Attention-RPN and Multi-Relation Detector

    论文地址：https://arxiv.org/abs/1908.01998

    AugFPN: Improving Multi-scale Feature Learning for Object Detection

    论文地址：https://arxiv.org/abs/1912.05384

    Hit-Detector: Hierarchical Trinity Architecture Search for Object Detection

    论文地址：https://arxiv.org/abs/2003.11818

    代码：https://github.com/ggjy/HitDet.pytorch

    Multi-task Collaborative Network for Joint Referring Expression Comprehension and Segmentation

    论文地址：https://arxiv.org/abs/2003.08813

    CentripetalNet: Pursuing High-quality Keypoint Pairs for Object Detection

    论文地址：https://arxiv.org/abs/2003.09119

    代码：https://github.com/KiveeDong/CentripetalNet

 

图像分割

    Semi-Supervised Semantic Image Segmentation with Self-correcting Networks

    论文地址：https://arxiv.org/abs/1811.07073

    Deep Snake for Real-Time Instance Segmentation

    论文地址：https://arxiv.org/abs/2001.01629

    CenterMask : Real-Time Anchor-Free Instance Segmentation

    论文地址：https://arxiv.org/abs/1911.06667

    代码：https://github.com/youngwanLEE/CenterMask

    SketchGCN: Semantic Sketch Segmentation with Graph Convolutional Networks

    论文地址：https://arxiv.org/abs/2003.00678

    PolarMask: Single Shot Instance Segmentation with Polar Representation

    论文地址：https://arxiv.org/abs/1909.13226

    代码：https://github.com/xieenze/PolarMask

    xMUDA: Cross-Modal Unsupervised Domain Adaptation for 3D Semantic Segmentation

    论文地址：https://arxiv.org/abs/1911.12676

    BlendMask: Top-Down Meets Bottom-Up for Instance Segmentation

    论文地址：https://arxiv.org/abs/2001.00309

    Enhancing Generic Segmentation with Learned Region Representations

    论文地址：https://arxiv.org/abs/1911.08564

 

人脸识别

    Towards Universal Representation Learning for Deep Face Recognition

    论文地址：https://arxiv.org/abs/2002.11841

    Suppressing Uncertainties for Large-Scale Facial Expression Recognition

    论文地址：https://arxiv.org/abs/2002.10392

    代码：https://github.com/kaiwang960112/Self-Cure-Network

    Face X-ray for More General Face Forgery Detection

    论文地址：https://arxiv.org/pdf/1912.13458.pdf

    Pose Agnostic Cross-spectral Hallucination via Disentangling Independent Factors

    论文地址：https://arxiv.org/abs/1909.04365

    Deep Spatial Gradient and Temporal Depth Learning for Face Anti-spoofing

    论文地址：https://arxiv.org/abs/2003.08061

    代码：https://github.com/clks-wzz/FAS-SGTD

    Learning Meta Face Recognition in Unseen Domains

    论文地址：https://arxiv.org/abs/2003.07733

    代码：https://github.com/cleardusk/MFR

 

目标跟踪

    ROAM: Recurrently Optimizing Tracking Model

    论文地址：https://arxiv.org/abs/1907.12006

 

三维点云&重建

    PF-Net: Point Fractal Network for 3D Point Cloud Completion

    论文地址：https://arxiv.org/abs/2003.00410

    PointAugment: an Auto-Augmentation Framework for Point Cloud Classification

    论文地址：https://arxiv.org/abs/2002.10876

    代码：https://github.com/liruihui/PointAugment/

    Learning multiview 3D point cloud registration

    论文地址：https://arxiv.org/abs/2001.05119

    C-Flow: Conditional Generative Flow Models for Images and 3D Point Clouds

    论文地址：https://arxiv.org/abs/1912.07009

    RandLA-Net: Efficient Semantic Segmentation of Large-Scale Point Clouds

    论文地址：https://arxiv.org/abs/1911.11236

    Total3DUnderstanding: Joint Layout, Object Pose and Mesh Reconstruction for Indoor Scenes from a Single Image

    论文地址：https://arxiv.org/abs/2002.12212

    Implicit Functions in Feature Space for 3D Shape Reconstruction and Completion

    论文地址：https://arxiv.org/abs/2003.01456

    In Perfect Shape: Certifiably Optimal 3D Shape Reconstruction from 2D Landmarks

    论文地址：https://arxiv.org/pdf/1911.11924.pdf

    Attentive Context Normalization for Robust Permutation-Equivariant Learning

    论文地址：https://arxiv.org/abs/1907.02545 Weiwei Sun, Wei Jiang, Eduard Trulls, Andrea Tagliasacchi, Kwang Moo Yi

    PQ-NET: A Generative Part Seq2Seq Network for 3D Shapes

    论文地址：https://arxiv.org/abs/1911.10949

    SG-NN: Sparse Generative Neural Networks for Self-Supervised Scene Completion of RGB-D Scans

    论文地址：https://arxiv.org/abs/1912.00036

    Cascade Cost Volume for High-Resolution Multi-View Stereo and Stereo Matching

    论文地址：https://arxiv.org/abs/1912.06378

    代码：https://github.com/alibaba/cascade-stereo

    Unsupervised Learning of Intrinsic Structural Representation Points

    论文地址：https://arxiv.org/abs/2003.01661

    代码：https://github.com/NolenChen/3DStructurePoints

 

图像处理

    Learning to Shade Hand-drawn Sketches

    论文地址：https://arxiv.org/abs/2002.11812

    Single Image Reflection Removal through Cascaded Refinement

    论文地址：https://arxiv.org/abs/1911.06634

    Generalized ODIN: Detecting Out-of-distribution Image without Learning from Out-of-distribution Data

    论文地址：https://arxiv.org/abs/2002.11297

    Deep Image Harmonization via Domain Verification

    论文地址：https://arxiv.org/abs/1911.13239

    代码：https://github.com/bcmi/Image_Harmonization_Datasets

    RoutedFusion: Learning Real-time Depth Map Fusion

    论文地址：https://arxiv.org/pdf/2001.04388.pdf

    Neural Contours: Learning to Draw Lines from 3D Shapes

    论文地址：https://arxiv.org/abs/2003.10333

    Towards Photo-Realistic Virtual Try-On by Adaptively Generating鈫Preserving Image Content

    论文地址：https://arxiv.org/abs/2003.05863

    Reinforced Feature Points: Optimizing Feature Detection and Description for a High-Level Task（图像处理-图像特征匹配）

    论文地址：https://arxiv.org/abs/1912.00623

    Correspondence Networks with Adaptive Neighbourhood Consensus（图像处理-图像特征匹配）

    论文地址：https://arxiv.org/abs/2003.12059

    Normalized and Geometry-Aware Self-Attention Network for Image Captioning（图像处理-图像字幕）

    论文地址：https://arxiv.org/abs/2003.08897

图像分类

    Self-training with Noisy Student improves ImageNet classification

    论文地址：https://arxiv.org/abs/1911.04252

    Image Matching across Wide Baselines: From Paper to Practice

    论文地址：https://arxiv.org/abs/2003.01587

    Towards Robust Image Classification Using Sequential Attention Models

    论文地址：https://arxiv.org/abs/1912.02184

    Learning in the Frequency Domain

    论文地址：https://arxiv.org/abs/2002.12416

    Learning from Web Data with Memory Module

    论文地址：https://arxiv.org/abs/1906.12028

    Making Better Mistakes: Leveraging Class Hierarchies with Deep Networks

    论文地址：https://arxiv.org/abs/1912.09393

 

### 姿态估计/动作识别

    VIBE: Video Inference for Human Body Pose and Shape Estimation

    论文地址：https://arxiv.org/abs/1912.05656

    代码：https://github.com/mkocabas/VIBE

    Distribution-Aware Coordinate Representation for Human Pose Estimation

    论文地址：https://arxiv.org/abs/1910.06278

    代码：https://github.com/ilovepose/DarkPose

    4D Association Graph for Realtime Multi-person Motion Capture Using Multiple Video Cameras

    论文地址：https://arxiv.org/abs/2002.12625

    Optimal least-squares solution to the hand-eye calibration problem

    论文地址：https://arxiv.org/abs/2002.10838

    D3VO: Deep Depth, Deep Pose and Deep Uncertainty for Monocular Visual Odometry

    论文地址：https://arxiv.org/abs/2003.01060

    Multi-Modal Domain Adaptation for Fine-Grained Action Recognition

    论文地址：https://arxiv.org/abs/2001.09691

    Distribution Aware Coordinate Representation for Human Pose Estimation

    论文地址：https://arxiv.org/abs/1910.06278

    The Devil is in the Details: Delving into Unbiased Data Processing for Human Pose Estimation

    论文地址：https://arxiv.org/abs/1911.07524

    PVN3D: A Deep Point-wise 3D Keypoints Voting Network for 6DoF Pose Estimation

    论文地址：https://arxiv.org/abs/1911.04231

    Action Segmentation with Joint Self-Supervised Temporal Domain Adaptation

    论文地址：https://arxiv.org/abs/2003.02824

    G2L-Net: Global to Local Network for Real-time 6D Pose Estimation with Embedding Vector Features

    论文地址：https://arxiv.org/abs/2003.11089

    Deep Image Spatial Transformation for Person Image Generation

    论文地址：https://arxiv.org/abs/2003.00696

    代码：https://github.com/RenYurui/ Global-Flow-Local-Attention

 

视频分析

    Rethinking Zero-shot Video Classification: End-to-end Training for Realistic Applications

    论文地址：https://arxiv.org/abs/2003.01455

    代码：https://github.com/bbrattoli/ZeroShotVideoClassification

    Say As You Wish: Fine-grained Control of Image Caption Generation with Abstract Scene Graphs

    论文地址：https://arxiv.org/abs/2003.00387

    Fine-grained Video-Text Retrieval with Hierarchical Graph Reasoning

    论文地址：https://arxiv.org/abs/2003.00392

    Object Relational Graph with Teacher-Recommended Learning for Video Captioning

    论文地址：https://arxiv.org/abs/2002.11566

    Zooming Slow-Mo: Fast and Accurate One-Stage Space-Time Video Super-Resolution

    论文地址：https://arxiv.org/abs/2002.11616

    Blurry Video Frame Interpolation

    论文地址：https://arxiv.org/abs/2002.12259

    Hierarchical Conditional Relation Networks for Video Question Answering

    论文地址：https://arxiv.org/abs/2002.10698

    Action Modifiers:Learning from Adverbs in Instructional Video

    论文地址：https://arxiv.org/abs/1912.06617

    Visual Grounding in Video for Unsupervised Word Translation

    论文地址：https://arxiv.org/abs/2003.05078

    代码：https://github.com/gsig/visual-grounding

    MaskFlownet: Asymmetric Feature Matching with Learnable Occlusion Mask（视频分析-光流估计）

    论文地址：https://arxiv.org/abs/2003.10955

    代码：https://github.com/microsoft/MaskFlownet

    Use the Force, Luke! Learning to Predict Physical Forces by Simulating Effects（视频预测）

    论文地址：https://arxiv.org/abs/2003.12045

    代码：https://ehsanik.github.io/forcecvpr2020

 

OCR

    ABCNet: Real-time Scene Text Spotting with Adaptive Bezier-Curve Network

    论文地址：https://arxiv.org/abs/2002.10200

    代码：https://github.com/Yuliang-Liu/bezier_curve_text_spotting,https://github.com/aim-uofa/adet

    Iterative Answer Prediction with Pointer-Augmented Multimodal Transformers for TextVQA

    论文地址：https://arxiv.org/abs/1911.06258

 

GAN

    Your Local GAN: Designing Two Dimensional Local Attention Mechanisms for Generative Models

    论文地址：https://arxiv.org/abs/1911.12287

    代码：https://github.com/giannisdaras/ylg

    MSG-GAN: Multi-Scale Gradient GAN for Stable Image Synthesis

    论文地址：https://arxiv.org/abs/1903.06048

    Robust Design of Deep Neural Networks against Adversarial Attacks based on Lyapunov Theory

    论文地址：https://arxiv.org/abs/1911.04636

    PSGAN: Pose and Expression Robust Spatial-Aware GAN for Customizable Makeup Transfer

    论文地址：https://arxiv.org/abs/1909.06956

 

小样本/零样本

    Improved Few-Shot Visual Classification

    论文地址：https://arxiv.org/pdf/1912.03432.pdf

    Meta-Transfer Learning for Zero-Shot Super-Resolution

    论文地址：https://arxiv.org/abs/2002.12213

    Instance Credibility Inference for Few-Shot Learning

    论文地址：https://arxiv.org/abs/2003.11853

    代码：https://github.com/Yikai-Wang/ICI-FSL

 

弱监督/无监督/自监督

    Rethinking the Route Towards Weakly Supervised Object Localization

    论文地址：https://arxiv.org/abs/2002.11359

    NestedVAE: Isolating Common Factors via Weak Supervision

    论文地址：https://arxiv.org/abs/2002.11576

    Unsupervised Reinforcement Learning of Transferable Meta-Skills for Embodied Navigation

    论文地址：https://arxiv.org/abs/1911.07450

    Disentangling Physical Dynamics from Unknown Factors for Unsupervised Video Prediction

    论文地址：https://arxiv.org/abs/2003.01460

    ClusterFit: Improving Generalization of Visual Representations

    论文地址：https://arxiv.org/abs/1912.03330

    Auto-Encoding Twin-Bottleneck Hashing

    论文地址：https://arxiv.org/abs/2002.11930

    Learning Representations by Predicting Bags of Visual Words

    论文地址：https://arxiv.org/abs/2002.12247

    A Characteristic Function Approach to Deep Implicit Generative Modeling

    论文地址：https://arxiv.org/abs/1909.07425

    Unsupervised Learning of Intrinsic Structural Representation Points

    论文地址：https://arxiv.org/abs/2003.01661

    代码：https://github.com/NolenChen/3DStructurePoints

 

行人跟踪/行人检测/ReID

    Cross-modality Person re-identification with Shared-Specific Feature Transfer

    论文地址：https://arxiv.org/abs/2002.12489

    Social-STGCNN: A Social Spatio-Temporal Graph Convolutional Neural Network for Human Trajectory Prediction

    论文地址：https://arxiv.org/abs/2002.11927

    The Garden of Forking Paths: Towards Multi-Future Trajectory Prediction

    论文地址：https://arxiv.org/abs/1912.06445

 

神经网络/模型压缩/模型加速

    GhostNet: More Features from Cheap Operations

    论文地址：https://arxiv.org/abs/1911.11907

    代码：https://github.com/iamhankai/ghostnet

    Watch your Up-Convolution: CNN Based Generative Deep Neural Networks are Failing to Reproduce Spectral

    论文地址：https://arxiv.org/abs/2003.01826

    GPU-Accelerated Mobile Multi-view Style Transfer

    论文地址：https://arxiv.org/abs/2003.00706

    Bundle Adjustment on a Graph Processor

    论文地址：https://arxiv.org/abs/2003.03134

    代码：https://github.com/joeaortiz/gbp

    Watch your Up-Convolution: CNN Based Generative Deep Neural Networks are Failing to Reproduce Spectral

    论文地址：https://arxiv.org/abs/2003.01826

    Holistically-Attracted Wireframe Parsing

    论文地址：https://arxiv.org/abs/2003.01663

    AdderNet: Do We Really Need Multiplications in Deep Learning?

    论文地址：https://arxiv.org/abs/1912.13200

    CARS: Contunuous Evolution for Efficient Neural Architecture Search

    论文地址：https://arxiv.org/abs/1909.04977

    代码：https://github.com/huawei-noah/CARS

    Π-nets: Deep Polynomial Neural Networksv

    论文地址：https://arxiv.org/abs/2003.03828

    Explaining Knowledge Distillation by Quantifying the Knowledge

    论文地址：https://arxiv.org/abs/2003.03622

 

超分辨率

    Zooming Slow-Mo: Fast and Accurate One-Stage Space-Time Video Super-Resolution

    论文地址：https://arxiv.org/abs/2002.11616

    Closed-loop Matters: Dual Regression Networks for Single Image Super-Resolution

    论文地址：https://arxiv.org/abs/2003.07018

    代码：https://github.com/guoyongcs/DRN

 

视觉常识/其他

    Visual Commonsense R-CNN

    论文地址：https://arxiv.org/abs/2002.12204

    代码：https://github.com/Wangt-CN/VC-R-CNN

    Scalable Uncertainty for Computer Vision with Functional Variational Inference

    论文地址：https://arxiv.org/abs/2003.03396

    Deep Representation Learning on Long-tailed Data: A Learnable Embedding Augmentation Perspective

    论文地址：https://arxiv.org/abs/2002.10826

    Representations, Metrics and Statistics For Shape Analysis of Elastic Graphs

    论文地址：https://arxiv.org/abs/2003.00287

    Filter Grafting for Deep Neural Networks

    论文地址：https://arxiv.org/abs/2001.05868

    代码：https://github.com/fxmeng/filter-grafting.git

    12-in-1: Multi-Task Vision and Language Representation Learning

    论文地址：https://arxiv.org/abs/1912.02315

    Towards Learning a Generic Agent for Vision-and-Language Navigation via Pre-training

    论文地址：https://arxiv.org/abs/2002.10638

    代码：https://github.com/weituo12321/PREVALENT

    Unbiased Scene Graph Generation from Biased Training

    论文地址：https://arxiv.org/abs/2002.11949

    Towards Visually Explaining Variational Autoencoders

    论文地址：https://arxiv.org/abs/1911.07389

    BBN: Bilateral-Branch Network with Cumulative Learning for Long-Tailed Visual Recognition

    论文地址：http://www.weixiushen.com/publication/cvpr20_BBN.pdf

    代码：https://github.com/Megvii-Nanjing/BBN

    High Frequency Component Helps Explain the Generalization of Convolutional Neural Networks

    论文地址：https://arxiv.org/abs/1905.13545

    SAM: The Sensitivity of Attribution Methods to Hyperparameters

    论文地址：http://s.anhnguyen.me/sam_cvpr2020.pdf

    代码：https://github.com/anguyen8/sam

    Π− nets: Deep Polynomial Neural Networks

    论文地址：https://arxiv.org/abs/2003.03828

    Towards Backward-Compatible Representation Learning

    论文地址：https://arxiv.org/abs/2003.11942

    On Translation Invariance in CNNs: Convolutional Layers can Exploit Absolute Spatial Location

    论文地址：https://arxiv.org/abs/2003.07064

    KeypointNet: A Large-scale 3D Keypoint Dataset Aggregated from Numerous Human Annotations（数据集）

    论文地址：https://arxiv.org/abs/2002.12687

————————————————
版权声明：本文为CSDN博主「极市平台」的原创文章，遵循CC 4.0 BY-SA版权协议，转载请附上原文出处链接及本声明。
原文链接：https://blog.csdn.net/Extremevision/article/details/104789697