【DG】 Survey DG Paper Reading list & Advice

in Artificial Intelligence

Survey DG papers

2.0. Existing 기법정리

Adaptive로 사용할 만한 Existing 기법들 / 핵심 문제점을 해결할 나만의 방법도 생각해보기

  1. Dynamic convolution / Conditional convolution
  2. Sparse-RCNN: proposal features convolution
  3. BlendMask -FCOS 기반
  4. YOLACT -RetinaNet 기반
  5. Memory and Centroid
  6. Attention (domain and shape)
  7. Critic (regularizer(self-sup 방법이면 충분?))
  8. Domain prototype (Embeding code)
  9. Generator (Manifold projector)

3.0. Paper list & Relative work

2020 이전 추천 논문 🌗

  1. (pass) Domain Randomization and Pyramid Consistency: Simulation-to-Real Generalization without Accessing Target Domain Data -ICCV18
  2. (pass) DLOW: Domain Flow for Adaptation and Generalization -CVPR19
  3. (pass) Learning from Extrinsic and Intrinsic Supervisions for Domain Generalization -ECCV20
  4. Domain Agnostic Learning with Disentangled Representations -ICML19
  5. (pass) ACE: Adapting to Changing Environments for Semantic Segmentation -ICCV19 (인규형 추천 스타일 메모리)
  6. Contrastive Syn-to-Real Generalization -ICLR21(인규형 추천)
  7. Unsupervised Domain Adaptation through Self-Supervision -ICLR20 (인규형 추천)
  8. Confidence Regularized Self-Training -ICCV19 (CRST)
  9. Domain Adaptation for Semantic Segmentation via Class-Balanced Self-Training -ECCV18 (CBST)

CVPR 21 paper list ⭐️

  1. (pass) FSDR: Frequency Space Domain Randomization for Domain Generalization
  2. (pass) Open Domain Generalization with Domain-Augmented Meta-Learning
  3. (pass) MOS- Towards Scaling Out-of-distribution Detection for Large Semantic Space
  4. (pass) MOOD- Multi-level Out-of-distribution Detection
  5. (pass) Prototypical Cross-domain Self-supervised Learning for Few-shot Unsupervised Domain Adaptation
  6. (pass) Domain-Irrelevant Representation Learning for Unsupervised Domain Generalization

4.0. Paper list & Relative work

[꼭 읽어야 하는 추천 Paper] ⭐️⭐️

  1. Source-Free Open Compound Domain Adaptation in Semantic Segmentation -CVPR21 (Source-free 공부)
  2. Adaptive Risk Minimization: A Meta-Learning Approach for Tackling Group Distribution Shift -ICLR21 (인규형 추천, 리뷰도 보기)
  3. Test-Time Adaptation to Distribution Shift by Confidence Maximization and Input Transformation -(Tent 후속 관용추천)
  4. Source-free domain adaptation for semantic segmentation via self-supervised selective self-training -(Tent 디스 영택 추천)
  5. Domain Adaptation for Semantic Segmentation with Maximum Squares Loss -ICCV19

[Adaptive/Dynamic Network DG papers] ⭐️

(Adaptive / Sementic Segmentation 을 키워드로 내가 사용할 수 있는 아이디어가 뭘까? 고민해보기!)

  1. Coarse-to-Fine Domain Adaptive Semantic Segmentation with Photometric Alignment and Category-Center Regularization -CVPR21
  2. Iterative Filter Adaptive Network for Single Image Defocus Deblurring -CVPR21
  3. Cross-View Regularization for Domain Adaptive Panoptic Segmentation -CVPR21
  4. Adaptive Aggregation Networks for Class-Incremental Learning -CVPR21
  5. Dynamic Transfer for Multi-Source Domain Adaptation -CVPR21
  6. Progressive Domain Expansion Network for Single Domain Generalization -CVPR21
  7. Cluster, Split, Fuse, and Update: Meta-Learning for Open Compound Domain Adaptive Semantic Segmentation -CVPR21

하지만 Test time / adaptive modules가 아닌 논문들 / 혹은 일단 패스..

  1. (pass) Adaptive Convolutions for Structure-Aware Style Transfer -CVPR21
  2. (pass) Dynamic Weighted Learning for Unsupervised Domain Adaptation -CVPR21

[공부할 Code/New DG paper]

  1. facebookresearch/DomainBed
  2. CBST / CRST 논문 및 코드 공부하기 (CRST 논문, CBST 논문)

[BaseLine 코드 찾기를 위한 사이트 모음]

  1. https://github.com/amusi/CVPR2021-Papers-with-Code
  2. https://github.com/52CV/CVPR-2021-Papers
  3. https://github.com/zhaoxin94/awesome-domain-adaptation
  4. https://github.com/amber0309/Domain-generalization

5.0. TENT 관련 논문 List

논문 검색방법

  1. arxiv, 구글스칼라, 구글 검색 등에 키워드 검색
  2. 그렇게 몇 논문이 나타나면 그 논문의 related work을 살펴보거나 그 논문을 인용한 논문을 재조사
  3. ICCV, CVPR에서 Source free 중심으로 논문 찾아보기

관련 논문 List

  1. Tent: Fully Test-time Adaptation by Entropy Minimization 20.6 ~ 21.5.18
  2. Adapting ImageNet-scale models to complex distribution shifts with self-learning 21.4.27
  3. Test-Time Adaptation to Distribution Shift by Confidence Maximization and Input Transformation - 21.6.28
  4. S4T: Source-free domain adaptation for semantic segmentation via self-supervised selective self-training - 21.7.21
  5. Generalized Source-free Domain Adaptation -21.8.3 (ICCV21)
  6. Compositional Models: Multi-Task Learning and Knowledge Transfer with Modular Networks -20.12~21.7.23 (ICLR 2021)
  7. SENTRY: Selective Entropy Optimization via Committee Consistency for Unsupervised Domain Adaptation -20.12.21

Notion 참고

© All rights reserved By Junha Song.