Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
158 tokens/sec
GPT-4o
7 tokens/sec
Gemini 2.5 Pro Pro
45 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
38 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

MLNet: Mutual Learning Network with Neighborhood Invariance for Universal Domain Adaptation (2312.07871v4)

Published 13 Dec 2023 in cs.CV

Abstract: Universal domain adaptation (UniDA) is a practical but challenging problem, in which information about the relation between the source and the target domains is not given for knowledge transfer. Existing UniDA methods may suffer from the problems of overlooking intra-domain variations in the target domain and difficulty in separating between the similar known and unknown class. To address these issues, we propose a novel Mutual Learning Network (MLNet) with neighborhood invariance for UniDA. In our method, confidence-guided invariant feature learning with self-adaptive neighbor selection is designed to reduce the intra-domain variations for more generalizable feature representation. By using the cross-domain mixup scheme for better unknown-class identification, the proposed method compensates for the misidentified known-class errors by mutual learning between the closed-set and open-set classifiers. Extensive experiments on three publicly available benchmarks demonstrate that our method achieves the best results compared to the state-of-the-arts in most cases and significantly outperforms the baseline across all the four settings in UniDA. Code is available at https://github.com/YanzuoLu/MLNet.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (46)
  1. Open Set Domain Adaptation. In ICCV, 754–763.
  2. Partial Adversarial Domain Adaptation. In ECCV, 135–150.
  3. Unified Optimal Transport Framework for Universal Domain Adaptation. In NeurIPS.
  4. Reusing the Task-specific Classifier as a Discriminator: Discriminator-free Adversarial Domain Adaptation. In CVPR, 7181–7190.
  5. Mutual Nearest Neighbor Contrast and Hybrid Prototype Self-Training for Universal Domain Adaptation. AAAI, 36(6): 6248–6257.
  6. Evidential Neighborhood Contrastive Learning for Universal Domain Adaptation. AAAI, 36(6): 6258–6267.
  7. Geometric Anchor Correspondence Mining with Uncertainty Modeling for Universal Domain Adaptation. In CVPR, 16134–16143.
  8. Universal Domain Adaptation from Foundation Models. arXiv:2305.11092.
  9. ImageNet: A large-scale hierarchical image database. In CVPR, 248–255.
  10. Adaptive Exploration for Unsupervised Person Re-identification. TOMM, 16(1): 1–19.
  11. Learning to Detect Open Classes for Universal Domain Adaptation. In ECCV, 567–583.
  12. Semi-supervised Learning by Entropy Minimization. In NeurIPS, volume 17.
  13. Deep Residual Learning for Image Recognition. In CVPR, 770–778.
  14. Learning Classifiers of Prototypes and Reciprocal Points for Universal Domain Adaptation. In WACV, 531–540.
  15. Towards Novel Target Discovery Through Open-Set Domain Adaptation. In ICCV, 9322–9331.
  16. A Review of Domain Adaptation without Target Labels. TPAMI, 43(3): 766–785.
  17. Universal Source-Free Domain Adaptation. In CVPR, 4544–4553.
  18. Deep learning. Nature, 521(7553): 436–444.
  19. Domain Consensus Clustering for Universal Domain Adaptation. In CVPR, 9757–9766.
  20. UMAD: Universal Model Adaptation under Domain and Category Shift. arXiv:2112.08553.
  21. Adversarial Partial Domain Adaptation by Cycle Inconsistency. In ECCV, 530–548.
  22. Learning to Adapt Across Dual Discrepancy for Cross-Domain Person Re-Identification. TPAMI, 45(2): 1963–1980.
  23. Virtual Mixup Training for Unsupervised Domain Adaptation. arXiv:1905.04215.
  24. Classmix: Segmentation-based data augmentation for semi-supervised learning. In WACV, 1369–1378.
  25. VisDA: The Visual Domain Adaptation Challenge. arXiv:1710.06924.
  26. Upcycling models under domain and category shift. In CVPR, 20019–20028.
  27. A Closer Look at Smoothness in Domain Adversarial Training. In ICML, 18378–18399.
  28. Adapting Visual Category Models to New Domains. In ECCV, 213–226.
  29. Universal Domain Adaptation through Self Supervision. In NeurIPS, volume 33, 16282–16292.
  30. OVANet: One-vs-All Network for Universal Domain Adaptation. In ICCV, 9000–9009.
  31. Open Set Domain Adaptation by Backpropagation. In ECCV, 153–168.
  32. Grad-cam: Visual explanations from deep networks via gradient-based localization. In ICCV, 618–626.
  33. Collaborative Learning of Diverse Experts for Source-free Universal Domain Adaptation. In ACM MM, 2054–2065.
  34. Neighborhood-based credibility anchor learning for universal domain adaptation. Pattern Recognition, 142: 109686.
  35. DACS: Domain Adaptation via Cross-domain Mixed Sampling. In WACV, 1379–1389.
  36. Visualizing data using t-SNE. Journal of Machine Learning Research, 9(11).
  37. Deep Hashing Network for Unsupervised Domain Adaptation. In CVPR, 5018–5027.
  38. Manifold Mixup: Better Representations by Interpolating Hidden States. In ICML, 6438–6447.
  39. Dual Mixup Regularized Learning for Adversarial Domain Adaptation. In ECCV, 540–555.
  40. Unsupervised Feature Learning via Non-parametric Instance Discrimination. In CVPR, 3733–3742.
  41. Improve Unsupervised Domain Adaptation with Mixup Training. arXiv:2001.00677.
  42. Deep Co-Training with Task Decomposition for Semi-Supervised Domain Adaptation. In ICCV, 8906–8916.
  43. One Ring to Bring Them All: Towards Open-Set Recognition under Domain Shift. arXiv:2206.03600.
  44. Universal Domain Adaptation. In CVPR, 2720–2729.
  45. mixup: Beyond Empirical Risk Minimization. In ICLR.
  46. Invariance Matters: Exemplar Memory for Domain Adaptive Person Re-Identification. In CVPR, 598–607.
Citations (8)
View on Semantic Scholar

Summary

We haven't generated a summary for this paper yet.

Ai Sparkles Streamline Icon: https://streamlinehq.com Summarize Now