README.md

CVPR2021 Workshop 组会最新论文/代码(持续更新)

🌟CVPR2021最新信息及已接收论文/代码(持续更新)

🎆🎆🎆更新提示：6月8日新增2篇

• Reducing the feature divergence of RGB and near-infrared images using Switchable Normalization
• 元学习
  • DAMSL: Domain Agnostic Meta Score-based Learning

🎆🎆🎆更新提示：6月7日新增1篇

• 细粒度
  • Fine-Grained Visual Classification of Plant Species In The Wild: Object Detection as A Reinforced Means of Attention

🎆🎆🎆更新提示：6月4日新增2篇

• Transformer
  • Anticipative Video Transformer
    🏠project
    在 CVPR 21 EPIC-Kitchens 行动预期挑战排行榜上排名第一
• 图像处理
  • NTIRE 2021 Challenge on High Dynamic Range Imaging: Dataset, Methods and Results

🎆🎆🎆更新提示：6月3日新增3篇

• 半监督
  • The Semi-Supervised iNaturalist Challenge at the FGVC8 Workshop
• 分割
  • Rethinking Cross-modal Interaction from a Top-down Perspective for Referring Video Object Segmentation
• 细粒度
  • Cleaning and Structuring the Label Space of the iMet Collection 2020
    ⭐code

🎆🎆🎆更新提示：6月2日新增3篇

• 语义分割
  • Detecting Anomalies in Semantic Segmentation with Prototypes
• 未分
  • PanoDR: Spherical Panorama Diminished Reality for Indoor Scenes
  • Semi-Supervised Disparity Estimation with Deep Feature Reconstruction

🎆🎆🎆更新提示：6月1日新增3篇

• 车辆
  • Connecting Language and Vision for Natural Language-Based Vehicle Retrieval
    ⭐code
• 目标检测
  • Training Domain-invariant Object Detector Faster with Feature Replay and Slow Learner
    ⭐code
    本文所介绍算法 A-NDFT，是对 NDFT 的改良版本。A-NDFT 利用两种加速技术，feature replay 和 slow learner。因此，在一个大规模的 UAVDT 基准上，它可以将 NDVT 的训练时间从 31 小时减少到 3 小时，同时仍然保持性能。
• 6D
  • Data-driven 6D Pose Tracking by Calibrating Image Residuals in Synthetic Domains
    ⭐code

🎆🎆🎆更新提示：5月31日新增1篇

• The Herbarium 2021 Half–Earth Challenge Dataset

🎆🎆🎆更新提示：5月28日新增1篇

• RSCA: Real-time Segmentation-based Context-Aware Scene Text Detection

🎆🎆🎆更新提示：5月27日新增1篇

• 计算成像
  • How to Calibrate Your Event Camera
    ⭐code

🎆🎆🎆更新提示：5月26日新增1篇

• 三维
  • Real-time Monocular Depth Estimation with Sparse Supervision on Mobile

🐱	🐶	🐭	🐹	🐯
38.Transformer	37.6D	36.OCR
35.Data Augmentation(数据增广)	34.Computational Photography(光学、几何、光场成像、计算摄影)	33.GAN	32.手语识别	31.图像分类
30.目标跟踪	29.Auto-ML&NAS	28.医学影像	27.人体姿态估计	26.无监督
25.SLAM/AR/VR/机器人	24.模型压缩&应用部署	23.人脸	22.重建	21.视频
20.三维	19.光流	18.图像检索	17.动作检测识别	16.人员重识别
15.遥感航空影像	14VQA	13.SR	12.图像分割	11.图像处理
10.目标检测	9.姿态估计	8.Camera Trap Images-相机陷阱图像	7.图像到图像翻译	6.手绘草图
5.车辆车牌与智能驾驶	4.数据集	3.各种神经网络	2.算法学习	1.Unkown(未分)

38.Transformer

• Anticipative Video Transformer
  🏠project
  在 CVPR 21 EPIC-Kitchens 行动预期挑战排行榜上排名第一

37.6D

• Data-driven 6D Pose Tracking by Calibrating Image Residuals in Synthetic Domains
  ⭐code

36.OCR

• 场景文本识别
  • RSCA: Real-time Segmentation-based Context-Aware Scene Text Detection

35.Data Augmentation(数据增广)

• Wisdom for the Crowd: Discoursive Power in Annotation Instructions for Computer Vision

34.Computational Photography(光学、几何、光场成像、计算摄影)

• How to Calibrate Your Event Camera
  ⭐code
• HDR 成像
  • ADNet: Attention-guided Deformable Convolutional Network for High Dynamic Range Imaging
    ⭐code

33.GAN

• Learning to Generate Novel Scene Compositions from Single Images and Videos
• Directional GAN: A Novel Conditioning Strategy for Generative Networks

32.手语识别

• ChaLearn LAP Large Scale Signer Independent Isolated Sign Language Recognition Challenge: Design, Results and Future Research

31.图像分类

• Boosting Co-teaching with Compression Regularization for Label Noise
  ⭐code
• 多标签分类
  • PLM: Partial Label Masking for Imbalanced Multi-label Classification
• 细粒度
  • Cleaning and Structuring the Label Space of the iMet Collection 2020
    ⭐code
  • Fine-Grained Visual Classification of Plant Species In The Wild: Object Detection as A Reinforced Means of Attention

30.目标跟踪

• Detecting and Matching Related Objects with One Proposal Multiple Predictions
• Differentiable Event Stream Simulator for Non-Rigid 3D Tracking
  🏠project
• City-Scale Multi-Camera Vehicle Tracking Guided by Crossroad Zones
  ⭐code

29.Auto-ML&NAS

• Auto-ML
  • Network Space Search for Pareto-Efficient Spaces

28.Medical Imaging医学影像

• GAN-Based Data Augmentation and Anonymization for Skin-Lesion Analysis: A Critical Review
• 医学图像识别
  • Can self-training identify suspicious ugly duckling lesions?
• 无监督检测
  • Unsupervised Detection of Cancerous Regions in Histology Imagery using Image-to-Image Translation

27.人体姿态估计

• Table Tennis Stroke Recognition Using Two-Dimensional Human Pose Estimation

26.无监督/半监督

• 无监督
  • Perceptual Loss for Robust Unsupervised Homography Estimation
• 半监督
  • The Semi-Supervised iNaturalist Challenge at the FGVC8 Workshop

25.SLAM/AR/VR/机器人

• Comparing Representations in Tracking for Event Camera-based SLAM
  ⭐code

24.Quantization/Pruning/Knowledge Distillation/Model Compression(量化、剪枝、蒸馏、模型压缩/扩展与优化)

• BasisNet: Two-stage Model Synthesis for Efficient Inference
• 知识蒸馏
  • Distill on the Go: Online knowledge distillation in self-supervised learning
• 量化
  • Do All MobileNets Quantize Poorly? Gaining Insights into the Effect of Quantization on Depthwise Separable Convolutional Networks Through the Eyes of Multi-scale Distributional Dynamics
  • Pareto-Optimal Quantized ResNet Is Mostly 4-bit
    ⭐code

23.Face人脸

• 人脸表情识别
  • I Only Have Eyes for You: The Impact of Masks On Convolutional-Based Facial Expression Recognition
• 人脸识别
  • EQFace: A Simple Explicit Quality Network for Face Recognition
    ⭐code

22.Reconstruction重建

• 3D 人体重建
  • Temporal Consistency Loss for High Resolution Textured and Clothed 3DHuman Reconstruction from Monocular Video

21.Video视频

• 视频恢复
  • Restoration of Video Frames from a Single Blurred Image with Motion Understanding
• 异常检测
  • An Efficient Approach for Anomaly Detection in Traffic Videos
  • Good Practices and A Strong Baseline for Traffic Anomaly Detection
    在 CVPR 2021 NVIDIA AI CITY 挑战赛中的 Traffic Anomaly Detection(交通异常检测)中排名第一
• 风格迁移
  • Automatic Non-Linear Video Editing Transfer

20.3D三维

• OmniLayout: Room Layout Reconstruction from Indoor Spherical Panoramas
  ⭐code
• 深度估计
  • Real-time Monocular Depth Estimation with Sparse Supervision on Mobile

19.Optical Flow光流

• OmniFlow: Human Omnidirectional Optical Flow
  🌻dataset

18.Image Retrieval图像检索

• Continual learning in cross-modal retrieval

17.Action Detection and Recognition动作检测识别

• action spotting-重点动作识别
  • Temporally-Aware Feature Pooling for Action Spotting in Soccer Broadcasts
  • Camera Calibration and Player Localization in SoccerNet-v2 and Investigation of their Representations for Action Spotting
    🌻dataset
• 动作检测
  • Three-stream network for enriched Action Recognition

16.Person Re-Identifications人员重识别

• Graph-based Person Signature for Person Re-Identifications
• 行人检测
  • Generalizable Multi-Camera 3D Pedestrian Detection
• 基于视频的 Reid
  • Video-based Person Re-identification without Bells and Whistles
    ⭐code

15.Aeria/Drones/Satellite/RS Image(航空影像/无人机)

• 三维重建
  • Machine-learned 3D Building Vectorization from Satellite Imagery

14VQA-视觉问答

• Dealing with Missing Modalities in the Visual Question Answer-Difference Prediction Task through Knowledge Distillation

13.SR-超分辨率

• 视频超分辨率
  • Efficient Space-time Video Super Resolution using Low-Resolution Flow and Mask Upsampling
  • NTIRE 2021 Challenge on Video Super-Resolution
    🏠project
• 图像超分辨率
  • Anchor-based Plain Net for Mobile Image Super-Resolution
    ⭐code

12.Image Segmentation图像分割

• 语义分割
  • Improving Online Performance Prediction for Semantic Segmentation
  • Plants Don't Walk on the Street: Common-Sense Reasoning for Reliable Semantic Segmentation
  • Rethinking Ensemble-Distillation for Semantic Segmentation Based Unsupervised Domain Adaptation
  • Detecting Anomalies in Semantic Segmentation with Prototypes
• 实例分割
  • Fashion-Guided Adversarial Attack on Person Segmentation
• 视频目标分割
  • Rethinking Cross-modal Interaction from a Top-down Perspective for Referring Video Object Segmentation

11.Image Processing图像处理

• NTIRE 2021 Challenge on High Dynamic Range Imaging: Dataset, Methods and Results
• 去除滤镜
  • Instagram Filter Removal on Fashionable Images
• 去雾
  • A Two-branch Neural Network for Non-homogeneous Dehazing via Ensemble Learning
    ⭐code
• 图像压缩
  • DANICE: Domain adaptation without forgetting in neural image compression
• 图像质量评估
  • Region-Adaptive Deformable Network for Image Quality Assessment
    ⭐code
  • Perceptual Image Quality Assessment with Transformers
    ⭐code
    在NTIRE 2021年感知IQA挑战中获得第一名
• 去雨
  • Multi-Scale Hourglass Hierarchical Fusion Network for Single Image Deraining
• 照片补光
  • NTIRE 2021 Depth Guided Image Relighting Challenge
    ⭐code
• 去模糊
  • NTIRE 2021 Challenge on Image Deblurring
    🏠project
• 图像补光
  • Multi-modal Bifurcated Network for Depth Guided Image Relighting
    ⭐[code](https://github.com/weitingchen83/NTIRE2021-Depth- Guided-Image-Relighting-MBNet)
    是 NTIRE 2021 深度指南一对一补光挑战赛的冠军
  • S3Net: A Single Stream Structure for Depth Guided Image Relighting
    ⭐code
    在 NTIRE 2021 深度引导的任意重新照明挑战中获得第3名
  • Physically Inspired Dense Fusion Networks for Relighting
    OIDDR-Net排名第二，AMIDR-Net 在 NTIRE 2021 年深度引导图像重光挑战中名列前五名
• 图像恢复
  • EDPN: Enhanced Deep Pyramid Network for Blurry Image Restoration
    ⭐code
• bokeh effect(背景虚化)
  • Stacked Deep Multi-Scale Hierarchical Network for Fast Bokeh Effect Rendering from a Single Image
    ⭐code

10.Object Detection目标检测

• LSPnet: A 2D Localization-oriented Spacecraft Pose Estimation Neural Network
• Pseudo-IoU: Improving Label Assignment in Anchor-Free Object Detection
  ⭐code
• Training Domain-invariant Object Detector Faster with Feature Replay and Slow Learner
  ⭐code
  本文所介绍算法 A-NDFT，是对 NDFT 的改良版本。A-NDFT 利用两种加速技术，feature replay 和 slow learner。因此，在一个大规模的 UAVDT 基准上，它可以将 NDVT 的训练时间从 31 小时减少到 3 小时，同时仍然保持性能。
• 3D目标检测
  • High-level camera-LiDAR fusion for 3D object detection with machine learning

9.Pose Estimation姿态估计

• Towards Automated and Marker-less Parkinson Disease Assessment: Predicting UPDRS Scores using Sit-stand videos

8.Camera Trap Images-相机陷阱图像

• Filtering Empty Camera Trap Images in Embedded Systems
  ⭐code

7.Image-to-Image Translation图像到图像翻译

• Dual Contrastive Learning for Unsupervised Image-to-Image Translation
  ⭐code

6.手绘草图

• On Training Sketch Recognizers for New Domains
• 工程草图生成
  • Engineering Sketch Generation for Computer-Aided Design

5.车辆车牌与智能驾驶

• 自动驾驶

  • MVFuseNet: Improving End-to-End Object Detection and Motion Forecasting through Multi-View Fusion of LiDAR Data
  • Multi-task Learning with Attention for End-to-end Autonomous Driving
  • End-to-End Interactive Prediction and Planning with Optical Flow Distillation for Autonomous Driving
    🏠project
  • Rethinking of Radar's Role: A Camera-Radar Dataset and Systematic Annotator via Coordinate Alignment
    🌻dataset

• 车辆重识别

  • A Strong Baseline for Vehicle Re-Identification
    ⭐code
  • An Empirical Study of Vehicle Re-Identification on the AI City Challenge
    ⭐code
    获得 CVPR 2021研讨会上，NVIDIA AI City Challenge（英伟达人工智能城市挑战赛）第2赛道（车辆重识别）的第一名。

• 车辆检索

  • SBNet: Segmentation-based Network for Natural Language-based Vehicle Search
    ⭐code
  • Connecting Language and Vision for Natural Language-Based Vehicle Retrieval
    ⭐code

4.Dataset数据集

• The Multi-Agent Behavior Dataset: Mouse Dyadic Social Interactions
• The iWildCam 2021 Competition Dataset
• GOO: A Dataset for Gaze Object Prediction in Retail Environments
  🌻dataset
• The Herbarium 2021 Half–Earth Challenge Dataset

3.各种神经网络

• CNN
  • AsymmNet: Towards ultralight convolution neural networks using asymmetrical bottlenecks
    ⭐code
• DNN
  • Fast Walsh-Hadamard Transform and Smooth-Thresholding Based Binary Layers in Deep Neural Networks
• BNN-二进制神经网络
  • A Bop and Beyond: A Second Order Optimizer for Binarized Neural Networks

2.算法学习

• 主动学习
  • A Mathematical Analysis of Learning Loss for Active Learning in Regression
• 对比学习
  • Contrastive Learning Improves Model Robustness Under Label Noise
• 类增量学习
  • Class-Incremental Learning with Generative Classifiers
    ⭐code
• 增量学习
  • IB-DRR: Incremental Learning with Information-Back Discrete Representation Replay
• 持续学习
  • Class-Incremental Experience Replay for Continual Learning under Concept Drift
• 联邦学习
  • Towards Fair Federated Learning with Zero-Shot Data Augmentation
  • Cluster-driven Graph Federated Learning over Multiple Domains
• 元学习
  • DAMSL: Domain Agnostic Meta Score-based Learning

1.Unkown未分

• Reconsidering CO2 emissions from Computer Vision
• Assessment of deep learning based blood pressure prediction from PPG and rPPG signals
• I Find Your Lack of Uncertainty in Computer Vision Disturbing
• Revisiting The Evaluation of Class Activation Mapping for Explainability: A Novel Metric and Experimental Analysis
• Patch Shortcuts: Interpretable Proxy Models Efficiently Find Black-Box Vulnerabilities
• The 5th AI City Challenge
• Width Transfer: On the (In)variance of Width Optimization
• Sign Segmentation with Changepoint-Modulated Pseudo-Labelling
• CASSOD-Net: Cascaded and Separable Structures of Dilated Convolution for Embedded Vision Systems and Applications
• Feedback control of event cameras
• Effectively Leveraging Attributes for Visual Similarity
• Dynamic-OFA: Runtime DNN Architecture Switching for Performance Scaling on Heterogeneous Embedded Platforms
  与最先进的技术相比，在Jetson Xavier NX 上使用 ImageNet 的实验结果表明，在相似的 ImageNet Top-1 精度下，该方法的速度最高可达 3.5倍（CPU），2.4倍（GPU），或者在相似的延迟下，精度更高 3.8%（CPU），5.1%（GPU）。
• High-Resolution Complex Scene Synthesis with Transformers
• Deep Graphics Encoder for Real-Time Video Makeup Synthesis from Example
• Texture Generation with Neural Cellular Automata
  🏠project
• Single View Geocentric Pose in the Wild
  ⭐code
• PAL: Intelligence Augmentation using Egocentric Visual Context Detection
• PanoDR: Spherical Panorama Diminished Reality for Indoor Scenes
• Semi-Supervised Disparity Estimation with Deep Feature Reconstruction
• Reducing the feature divergence of RGB and near-infrared images using Switchable Normalization
• 异常检测
  • Brittle Features May Help Anomaly Detection