binli123 / dsmil-wsi

DSMIL: Dual-stream multiple instance learning networks for tumor detection in Whole Slide Image
MIT License
358 stars 88 forks source link

Precomputed features #72

Closed gadermayr closed 6 months ago

gadermayr commented 1 year ago

I have a question regarding the precomputed features (Camelion/TCGA) which can be downloaded (e.g. $ python download.py --dataset=tcga).

Where these features generated with a ResNet trained on image-net, or based on self supervised training (simclr)? Are the features for the other case also available in the precomputed form?

Thank you very much!

binli123 commented 1 year ago

Those are based on self-supervised training.

gadermayr commented 1 year ago

Thank you very much for the information!

jan-engelmann commented 1 year ago

thank you! And are these features multi scale or single scale for Camelion?

jan-engelmann commented 1 year ago

found the answer here: https://github.com/binli123/dsmil-wsi/issues/50