CVPR 2021 accepted papers on Label-efficient and robust learning
267 of the 1,661 papers accepted at CVPR 2021 (16.1%) are on Label-efficient and robust learning, compared with 14.9% at CVPR 2020. They fall into 9 topics.
See CVPR 2021 in the interactive map
Submitting to CVPR? Checklist →
Working on something in this area? Describe your idea in scime atlas to see the closest papers by meaning and how crowded the spot has become.
Topics and papers
Recognition and long-tail · 113 papers
- MagFace: A Universal Representation for Face Recognition and Quality Assessment (617 citations)
- Understanding the Behaviour of Contrastive Loss (615 citations)
- Patch-NetVLAD: Multi-Scale Fusion of Locally-Global Descriptors for Place Recognition (442 citations)
- Diverse Part Discovery: Occluded Person Re-Identification With Part-Aware Transformer (421 citations)
- CFNet: Cascade and Fused Cost Volume for Robust Stereo Matching (350 citations)
- General Multi-Label Image Classification With Transformers (349 citations)
- Frequency-Aware Discriminative Feature Learning Supervised by Single-Center Loss for Face Forgery Detection (324 citations)
- Discover Cross-Modality Nuances for Visible-Infrared Person Re-Identification (316 citations)
- Contrastive Learning Based Hybrid Networks for Long-Tailed Image Classification (298 citations)
- Deep Stable Learning for Out-of-Distribution Generalization (283 citations)
- Improving Calibration for Long-Tailed Recognition (281 citations)
- WebFace260M: A Benchmark Unveiling the Power of Million-Scale Deep Face Recognition (278 citations)
and 101 more in the interactive map.
Domain adaptation and generalization · 59 papers
- A Fourier-Based Framework for Domain Generalization (561 citations)
- Prototypical Pseudo Label Denoising and Target Structure Learning for Domain Adaptive Semantic Segmentation (541 citations)
- Reducing Domain Gap by Reducing Style Bias (351 citations)
- RobustNet: Improving Domain Generalization in Urban-Scene Segmentation via Instance Selective Whitening (326 citations)
- FSDR: Frequency Space Domain Randomization for Domain Generalization (286 citations)
- Source-Free Domain Adaptation for Semantic Segmentation (273 citations)
- Learning to Generalize Unseen Domains via Memory-based Multi-Source Meta-Learning for Person Re-Identification (244 citations)
- FixBi: Bridging Domain Spaces for Unsupervised Domain Adaptation (232 citations)
- Few-Shot Image Generation via Cross-Domain Correspondence (226 citations)
- Group-aware Label Transfer for Domain Adaptive Person Re-identification (214 citations)
- Progressive Domain Expansion Network for Single Domain Generalization (195 citations)
- Cross-Domain Gradient Discrepancy Minimization for Unsupervised Domain Adaptation (188 citations)
and 47 more in the interactive map.
Semi-supervised learning · 23 papers
- Semi-Supervised Semantic Segmentation With Cross Pseudo Supervision (1,204 citations)
- Meta Pseudo Labels (520 citations)
- CReST: A Class-Rebalancing Self-Training Framework for Imbalanced Semi-Supervised Learning (284 citations)
- Semantic Segmentation With Generative Models: Semi-Supervised Learning and Strong Out-of-Domain Generalization (173 citations)
- SimPLE: Similar Pseudo Label Exploitation for Semi-Supervised Classification (168 citations)
- Memory Oriented Transfer Learning for Semi-Supervised Image Deraining (118 citations)
- Adaptive Consistency Regularization for Semi-Supervised Transfer Learning (117 citations)
- Exponential Moving Average Normalization for Self-Supervised and Semi-Supervised Learning (117 citations)
- OpenMix: Reviving Known Knowledge for Discovering Novel Visual Categories in an Open World (107 citations)
- What if We Only Use Real Datasets for Scene Text Recognition? Toward Scene Text Recognition With Fewer Labels (105 citations)
- Three Ways To Improve Semantic Segmentation With Self-Supervised Depth Estimation (93 citations)
- ATSO: Asynchronous Teacher-Student Optimization for Semi-Supervised Image Segmentation (68 citations)
and 11 more in the interactive map.
Self-supervised and contrastive · 21 papers
- Exploring Simple Siamese Representation Learning (3,493 citations)
- Dual-Stream Multiple Instance Learning Network for Whole Slide Image Classification With Self-Supervised Contrastive Learning (888 citations)
- Self-Supervised Augmentation Consistency for Adapting Semantic Segmentation (244 citations)
- Pareto Self-Supervised Training for Few-Shot Learning (147 citations)
- The Spatially-Correlative Loss for Various Image Translation Tasks (126 citations)
- AdCo: Adversarial Contrast for Efficient Learning of Unsupervised Representations From Self-Trained Negative Adversaries (117 citations)
- Robust Audio-Visual Instance Discrimination (109 citations)
- OBoW: Online Bag-of-Visual-Words Generation for Self-Supervised Learning (75 citations)
- Multiple Instance Captioning: Learning Representations From Histopathology Textbooks and Articles (70 citations)
- Jigsaw Clustering for Unsupervised Visual Representation Learning (69 citations)
- The Lottery Tickets Hypothesis for Supervised and Self-Supervised Pre-Training in Computer Vision Models (69 citations)
- Learning a Self-Expressive Network for Subspace Clustering (68 citations)
and 9 more in the interactive map.
Zero-/few-shot and meta-learning · 16 papers
- Few-Shot Classification With Feature Map Reconstruction Networks (358 citations)
- Contrastive Embedding for Generalized Zero-Shot Learning (286 citations)
- Counterfactual Zero-Shot and Open-Set Visual Recognition (199 citations)
- Few-Shot Segmentation Without Meta-Learning: A Good Transductive Inference Is All You Need? (188 citations)
- Prototype Completion With Primitive Knowledge for Few-Shot Learning (162 citations)
- Learning Dynamic Alignment via Meta-Filter for Few-Shot Learning (134 citations)
- Exploring Complementary Strengths of Invariant and Equivariant Representations for Few-Shot Learning (133 citations)
- Light Field Super-Resolution With Zero-Shot Learning (73 citations)
- Rethinking Class Relations: Absolute-Relative Supervised and Unsupervised Few-Shot Learning (71 citations)
- ECKPN: Explicit Class Knowledge Propagation Network for Transductive Few-Shot Learning (68 citations)
- Few-Shot Open-Set Recognition by Transformation Consistency (62 citations)
- Fine-Grained Angular Contrastive Learning With Coarse Labels (54 citations)
and 4 more in the interactive map.
Noisy labels · 14 papers
- Jo-SRC: A Contrastive Approach for Combating Noisy Labels (183 citations)
- Partially View-Aligned Representation Learning With Noise-Robust Contrastive Loss (172 citations)
- Multi-Objective Interpolation Training for Robustness To Label Noise (143 citations)
- Augmentation Strategies for Learning With Noisy Labels (113 citations)
- Re-Labeling ImageNet: From Single to Multi-Labels, From Global to Localized Labels (109 citations)
- A Second-Order Approach to Learning With Instance-Dependent Label Noise (85 citations)
- Joint Negative and Positive Learning for Noisy Labels (73 citations)
- Faster Meta Update Strategy for Noise-Robust Deep Learning (63 citations)
- ProSelfLC: Progressive Self Label Correction for Training Robust Deep Neural Networks (63 citations)
- Noise-Resistant Deep Metric Learning With Ranking-Based Instance Selection (38 citations)
- DualGraph: A Graph-Based Method for Reasoning About Label Noise (37 citations)
- Correlated Input-Dependent Label Noise in Large-Scale Image Classification (36 citations)
and 2 more in the interactive map.
Knowledge distillation · 14 papers
- Distilling Knowledge via Knowledge Review (568 citations)
- Distilling Object Detectors via Decoupled Features (263 citations)
- General Instance Distillation for Object Detection (259 citations)
- Complementary Relation Contrastive Distillation (110 citations)
- SuperMix: Supervising the Mixing Data Augmentation (106 citations)
- Wasserstein Contrastive Representation Distillation (92 citations)
- EvDistill: Asynchronous Events To End-Task Learning via Bidirectional Reconstruction-Guided Cross-Modal Knowledge Distillation (81 citations)
- Fair Feature Distillation for Visual Recognition (56 citations)
- Learning Student Networks in the Wild (31 citations)
- Asymmetric Metric Learning for Knowledge Transfer (30 citations)
- Revisiting Knowledge Distillation: An Inheritance and Exploration Framework (29 citations)
- Probabilistic Model Distillation for Semantic Correspondence (24 citations)
and 2 more in the interactive map.
OOD and anomaly detection · 6 papers
- CutPaste: Self-Supervised Learning for Anomaly Detection and Localization (1,006 citations)
- PANDA: Adapting Pretrained Features for Anomaly Detection and Segmentation (285 citations)
- MOS: Towards Scaling Out-of-Distribution Detection for Large Semantic Space (159 citations)
- MOOD: Multi-Level Out-of-Distribution Detection (78 citations)
- Out-of-Distribution Detection Using Union of 1-Dimensional Subspaces (56 citations)
- Multiresolution Knowledge Distillation for Anomaly Detection
Test-time adaptation · 1 paper
- Test-Time Fast Adaptation for Dynamic Scene Deblurring via Meta-Auxiliary Learning (72 citations)
