【论文推送】跨域学习、跨域小样本学习Cross-Domain(持续更新...)

    深度神经网络(DNN)的泛化性能相当依赖于训练集的规模(size)和丰富性(variations). 但是在很多场景下无法搜集到如此多的数据, 如皮肤病, 卫星图片(飞机残骸). 尽管每种特殊的情况发生的概率很低, 人为处理的成本也很低, 但当我们有大量的”特殊情况”时, 问题就变得复杂了起来.

    如何利用一个或一些成熟域上的先验知识，去解决另一个样缺少样本域上的复杂任务是跨域少样本学习 (Cross-Domain Few Shot Learning)研究的主要问题。

本文主要记录该领域值得一看的文章，后面有机会也会陆续将笔记整理，并发布于：

AI研修 潜水摸大鱼

 

综述与数据集类：

1. Methodologies for cross-domain data fusion: An overview

2. A broader study of cross-domain few-shot learning

3. A closer look at few-shot classification

2021：

1. Prototypical Cross-domain Self-supervised Learning for Few-shot Unsupervised Domain Adaptation.

2. Adaptive Cross-Modal Prototypes for Cross-Domain Visual-Language Retrieval.

3. OTCE: A Transferability Metric for Cross-Domain Cross-Task Representations.

4. DiCyc: GAN-based deformation invariant cross-domain information fusion for medical image synthesis.

5. Generative self-training for cross-domain unsupervised tagged-to-cine mri synthesis

6. Is In-Domain Data Really Needed? A Pilot Study on Cross-Domain Calibration for Network Quantization

7. Exploring Stylometric and Emotion-Based Features for Multilingual Cross-Domain Hate Speech Detection

8. Informative and Consistent Correspondence Mining for Cross-Domain Weakly Supervised Object Detection

9. Learning deep semantic segmentation network under multiple weakly-supervised constraints for cross-domain remote sensing image semantic segmentation

10. Cross-Domain Gradient Discrepancy Minimization for Unsupervised Domain Adaptation

11. Tracklet-Refined Multi-Camera Tracking Based on Balanced Cross-Domain Re-Identification for Vehicles

12. DACS: Domain adaptation via cross-domain mixed sampling

13. Few-shot Image Generation via Cross-domain Correspondence

14. Cross-Domain Adaptive Clustering for Semi-Supervised Domain Adaptation

15. Cross-Domain Multi-Task Learning for Object Detection and Saliency Estimation

16. Error-aware density isomorphism reconstruction for unsupervised cross-domain crowd counting

17. Domain structure-based transfer learning for cross-domain word representation

18. Cross-Domain Similarity Learning for Face Recognition in Unseen Domains

19. Face Parsing From RGB and Depth Using Cross-Domain Mutual Learning

20. Unbiased Mean Teacher for Cross-Domain Object Detection

21. Explanation-guided training for cross-domain few-shot classification

22. DRANet: Disentangling Representation and Adaptation Networks for Unsupervised Cross-Domain Adaptation

2020：

1. Self-challenging improves cross-domain generalization

2. Graph optimal transport for cross-domain alignment

3. Cross-domain detection via graph-induced prototype alignment

4. Cross-domain correspondence learning for exemplar-based image translation

5. Cross-domain object detection through coarse-to-fine feature adaptation

6. A cross-domain approach to analyzing the short-run impact of COVID-19 on the US electricity sector

7. FakeCovid--A multilingual cross-domain fact check news dataset for COVID-19

8. Joint disentangling and adaptation for cross-domain person re-identification

9. A broader study of cross-domain few-shot learning

10. Transferring cross-domain knowledge for video sign language recognition

11. Classes matter: A fine-grained adversarial approach to cross-domain semantic segmentation

12. On the limits of cross-domain generalization in automated X-ray prediction

13. Cross-domain semantic segmentation via domain-invariant interactive relation transfer

14. Cross-domain few-shot classification via learned feature-wise transformation

15. Cross-domain face presentation attack detection via multi-domain disentangled representation learning

16. Smoothing adversarial domain attack and p-memory reconsolidation for cross-domain person re-identification

17. BuildSenSys: Reusing Building Sensing Data for Traffic Prediction With Cross-Domain Learning

18. Ddtcdr: Deep dual transfer cross domain recommendation

19. Asymmetric co-teaching for unsupervised cross-domain person re-identification

20. Adversarial cross-domain action recognition with co-attention

21. Coach: A coarse-to-fine approach for cross-domain slot filling

22. Lcd: Learned cross-domain descriptors for 2d-3d matching

2019：

1. Crdoco: Pixel-level domain transfer with cross-domain consistency

2. Self-challenging improves cross-domain generalization

3. Adaptive transfer network for cross-domain person re-identification

4. Exploring object relation in mean teacher for cross-domain detection

5. Exploring object relation in mean teacher for cross-domain detection

6. Adapting object detectors via selective cross-domain alignment

7. Cross-domain detection via graph-induced prototype alignment

8. Cross-domain correspondence learning for exemplar-based image translation

9. Multi-level domain adaptive learning for cross-domain detection

10. Learning to discover cross-domain relations with generative adversarial networks

11. Cross-domain few-shot learning with meta fine-tuning

12. Cross-Domain Few-Shot Learning by Representation Fusion

其他：

1.  Explain and improve: Cross-domain few-shot-learning using explanations

2. A closer look at few-shot classification

3. Meta-FDMixup: Cross-Domain Few-Shot Learning Guided by Labeled Target Data

4. Few-shot learning via embedding adaptation with set-to-set functions

5. Few-shot acoustic event detection via meta learning

更多精彩内容：