search

binli123 / dsmil-wsi

DSMIL: Dual-stream multiple instance learning networks for tumor detection in Whole Slide Image
MIT License
332 stars 84 forks source link

patch threshold and entropy #65

Open shkeum21 opened 1 year ago

shkeum21 commented 1 year ago

Dear Binli.

First of all, I want to say that your paper and repository is contributing a lot to the development of WSI ml area. Your code implementation is much easier to understand and implement than others. Thanks :)

By the way, I'm little confused with the deepzoom_tiler.py args 'background_t' and your description in paper 'entropy'. IS filtering background with background_t = 15 same as entropy=5? Is there any formula for exchanging background_t and entropy?

Also, by using background_t = 15, i got more patches than expected. Would you offer us the exact background_t for tcga and camelyon? If it is confidential, just let us know the 'good' range of background_t for both datasets.

Have a good day :)

yogawicaksana commented 1 year ago

Hi, Have you figure it out this one? Is the background_t=15 same as entropy=5?

Thank you

shkeum21 commented 1 year ago

No, I'm just waiting for the reply from author @binli123