Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
125 tokens/sec
GPT-4o
53 tokens/sec
Gemini 2.5 Pro Pro
42 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
47 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Source-Free Domain Adaptation of Weakly-Supervised Object Localization Models for Histology (2404.19113v2)

Published 29 Apr 2024 in cs.CV and cs.LG

Abstract: Given the emergence of deep learning, digital pathology has gained popularity for cancer diagnosis based on histology images. Deep weakly supervised object localization (WSOL) models can be trained to classify histology images according to cancer grade and identify regions of interest (ROIs) for interpretation, using inexpensive global image-class annotations. A WSOL model initially trained on some labeled source image data can be adapted using unlabeled target data in cases of significant domain shifts caused by variations in staining, scanners, and cancer type. In this paper, we focus on source-free (unsupervised) domain adaptation (SFDA), a challenging problem where a pre-trained source model is adapted to a new target domain without using any source domain data for privacy and efficiency reasons. SFDA of WSOL models raises several challenges in histology, most notably because they are not intended to adapt for both classification and localization tasks. In this paper, 4 state-of-the-art SFDA methods, each one representative of a main SFDA family, are compared for WSOL in terms of classification and localization accuracy. They are the SFDA-Distribution Estimation, Source HypOthesis Transfer, Cross-Domain Contrastive Learning, and Adaptively Domain Statistics Alignment. Experimental results on the challenging Glas (smaller, breast cancer) and Camelyon16 (larger, colon cancer) histology datasets indicate that these SFDA methods typically perform poorly for localization after adaptation when optimized for classification.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (72)
  1. Grad-cam++: Generalized gradient-based visual explanations for deep convolutional networks. In WACV.
  2. Unsupervised robust domain adaptation without source data. CoRR, abs/2103.14577.
  3. Imagenet classification with deep convolutional neural networks. In NeurIps.
  4. Weakly supervised object localization via transformer with implicit spatial calibration. ECCV.
  5. F-CAM: Full resolution class activation maps via guided parametric upscaling. In WACV.
  6. Grad-cam++: Generalized gradient-based visual explanations for deep convolutional networks. In WACV, pages 839–847.
  7. Self-supervised noisy label learning for source-free unsupervised domain adaptation. CoRR, abs/2102.11614.
  8. LCTR: On awakening the local continuity of transformer for weakly supervised object localization. In AAAI, volume 36, pages 410–418.
  9. Attention-based dropout layer for weakly supervised object localization. In CVPR.
  10. Source-free domain adaptation via distribution estimation. CoRR, abs/2204.11257.
  11. Proxymix: Proxy-based mixup training with label refinery for source-free domain adaptation. CoRR, abs/2205.14566.
  12. Unsupervised domain adaptation by statistics alignment for deep sleep staging networks. IEEE TNSRE, 30:205–216.
  13. Source-free unsupervised domain adaptation: A survey. NEUNet, page 106230.
  14. A brief review of domain adaptation. ICDATA, pages 877–894.
  15. Fawcett, T. (2006). An introduction to roc analysis. Pattern recognition letters, 27(8):861–874.
  16. TS-CAM: Token semantic coupled attention map for weakly supervised object localization. In ICCV.
  17. TS-CAM: token semantic coupled attention map for weakly supervised object localization. In ICCV.
  18. Digital image processing, 3rd Edition. Pearson Education.
  19. ViTOL: Vision transformer for weakly supervised object localization. In CVPRw.
  20. Histology image analysis for carcinoma detection and grading. CMPB, 107(3):538–556.
  21. Uncertainty-aware mean teacher for source-free unsupervised domain adaptive 3d object detection. CoRR, abs/2109.14651.
  22. Source free domain adaptation with image translation. CoRR, abs/2008.07514.
  23. Source-free domain adaptation via distributional alignment by matching batch normalization statistics. CoRR, abs/2101.10842.
  24. Evaluating weakly supervised object localization methods right. In CVPR.
  25. Contrastive adaptation network for unsupervised domain adaptation. CoRR, abs/1901.00976.
  26. Machine learning methods for histopathological image analysis. CSBJ, 16:34–42.
  27. A review of domain adaptation without target labels. IEEE TPAMI, 43(3):766–785.
  28. Domain impression: A source data free domain adaptation method. CoRR, abs/2102.09003.
  29. Li, M. (2022). CaFT: Clustering and filter on tokens of transformer for weakly supervised object localization. CoRR, abs/2201.00475.
  30. Model adaptation: Unsupervised domain adaptation without source data. In CVPR, pages 9638–9647.
  31. Adapting off-the-shelf source segmenter for target medical image segmentation. In MICCAI 2021, pages 549–559, Cham. Springer International Publishing.
  32. A source-free domain adaptive polyp detection framework with style diversification flow. IEEE TMI, 41(7):1897–1908.
  33. Graph consistency based mean-teaching for unsupervised domain adaptive person re-identification. CoRR, abs/2105.04776.
  34. Attention-based deep multiple instance learning. In ICML.
  35. Foreground activation maps for weakly supervised object localization. In ICCV, pages 3385–3395.
  36. Adversarial transformers for weakly supervised object localization. IEEE TIP, 31:7130–7143.
  37. Constrained sampling for class-agnostic weakly supervised object localization. In Montreal AI symposium.
  38. Discriminative sampling of proposals in self-supervised transformers for weakly supervised object localization. CoRR, abs/2209.09209.
  39. Dips: Discriminative pseudo-label sampling with self-supervised transformers for weakly supervised object localization. IMAVIS, page 104838.
  40. Unveiling the potential of structure preserving for weakly supervised object localization. In CVPR, pages 11642–11651.
  41. Source-free domain adaptation via avatar prototype generation and adaptation. CoRR, abs/2106.15326.
  42. Pairwise similarity knowledge transfer for weakly supervised object localization. In ECCV.
  43. U-net: Convolutional networks for biomedical image segmentation. In MICCAI.
  44. Deep weakly-supervised learning methods for classification and localization in histology images: A survey. MLBI, 2:96–150.
  45. A survey of thresholding techniques. CVGIP, 41(2):233–260.
  46. Grad-cam: Visual explanations from deep networks via gradient-based localization. In ICCV, pages 618–626.
  47. Hide-and-seek: Forcing a network to be meticulous for weakly-supervised object and action localization. In ICCV.
  48. Re-attention transformer for weakly supervised object localization. In BMVC.
  49. Subtypes, R. (2022). Pathology of breast cancer. Updates in the Management of Breast Cancer, An Issue of Surgical Clinics, 103(1):1.
  50. Vdm-da: Virtual domain modeling for source data-free domain adaptation. CoRR, abs/2103.14357.
  51. Deep visual domain adaptation: A survey. NEUCOMPU, 312:135–153.
  52. Cross-domain contrastive learning for unsupervised domain adaptation. IEEE TMM.
  53. Cross-domain contrastive learning for unsupervised domain adaptation. CoRR, abs/2106.05528.
  54. Object region mining with adversarial erasing: A simple classification to semantic segmentation approach. In CVPR.
  55. Background activation suppression for weakly supervised object localization. In CVPR.
  56. Adaptive adversarial network for source-free domain adaptation. In ICCV, pages 8990–8999.
  57. Online refinement of low-level feature based activation map for weakly supervised object localization. In ICCV, pages 132–141.
  58. Danet: Divergent activation for weakly supervised object localization. In CVPR.
  59. Source free domain adaptation for medical image segmentation with fourier style mining. Medical Image Analysis, 79:102457.
  60. Transformer-based source-free domain adaptation. CoRR, abs/2105.14138.
  61. Exploiting the intrinsic neighborhood structure for source-free domain adaptation. NeurIPS, 34:29393–29405.
  62. Generalized source-free domain adaptation. In ICCV, pages 8978–8987.
  63. Source-free domain adaptation for real-world image dehazing. CoRR, abs/2207.06644.
  64. A comprehensive survey on source-free domain adaptation. CoRR, abs/2302.11803.
  65. Cutmix: Regularization strategy to train strong classifiers with localizable features. In ICCV, pages 6023–6032.
  66. Rethinking the route towards weakly supervised object localization. In CVPR.
  67. Source-free unsupervised domain adaptation: Current research and future directions. NERUCOMPU, page 126921.
  68. Adversarial complementary learning for weakly supervised object localization. In CVPR.
  69. Inter-image communication for weakly supervised localization. In ECCV, pages 271–287.
  70. Learning deep features for discriminative localization. In CVPR.
  71. Background-aware classification activation map for weakly supervised object localization. IEEE TPAMI.
  72. Weakly supervised object localization as domain adaption. In CVPR.
Citations (1)
View on Semantic Scholar

Summary

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

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