Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Negative patch oversampling #30

Open
6 of 8 tasks
4pygmalion opened this issue Jul 18, 2024 · 1 comment
Open
6 of 8 tasks

Negative patch oversampling #30

4pygmalion opened this issue Jul 18, 2024 · 1 comment
Assignees

Comments

@4pygmalion
Copy link
Owner

4pygmalion commented Jul 18, 2024

Summary

  • Stroma vs epithelium은 구분은 잘하는편
  • normal epithelium (N) vs lymphcytes (N) vs cancer은 구분을 못함

TODO

  • : Negative patch oversampling
    • : Custom dataset을 생성
    • : Custom dataset은 image, mask을 초기화시에 입력받아서, 각 이미지에 대해서 oversampling하는 방법을 만듬
    • : CustomDataSet._negative_oversample: 이라는 메서드에서 각 이미지를 돌면서 negative 영역이 큰 부분을 crop하여 random patch할 수 있도록 작성. 그리고 그 이미지는 다시 self.image = image에 할당해서 뒷단에서 해결
    • : 배경은 0이 아닌 255로 해결
  • : Cut and paste
  • : contrastive learning
  • : joint multi-task learning
@4pygmalion 4pygmalion self-assigned this Jul 18, 2024
4pygmalion added a commit that referenced this issue Jul 19, 2024
negative patch oversampling 추가 (issue #30)
@4pygmalion
Copy link
Owner Author

Color jittering으로 일부 성능 확인

  • Normal tissue가 보라색으로 염색되서, Cancer랑 헷갈려하는것 같음
  • 형태만으로 학습할 수 있도록 ToGray 및 Color jittering 방법추가

=> DICE: 0.811 (0.07%p) , IoU=0.686 (0.01%p) 향상

4pygmalion added a commit that referenced this issue Jul 26, 2024
4pygmalion added a commit that referenced this issue Jul 26, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants