Model sharing

[andreped]

Hello, @sophiajw!

Would it be possible to share the model and inference script to easily test and use the model?

I am developing an open-source platform called FastPathology, which enables researchers to easily deploy deep learning-based pipelines on their own data, without needing to code.

The software has been lacking a stain normalization method, and I am currently testing existing solutions to find out which to integrate. Macenko is my current bet, but as it is not as robust as more modern SOTA approaches, it would be great to get access to even better solutions. I have not been able to test any DL-methods, and none of these are public, but perhaps you could make that possible?

And of course, if the model is used in studies we will give appropriate credit.

[sophiajw]

hey andreped, sure! that sounds great!

Did you try the steps in the section 'Application' of the README?
I shared the link to the model weights trained on Camelyon17 there including the code that you should use to apply it as an augmentation in the forward pass of your downstream model.

I also have model weights trained on TCGA that should work better. Let me know when everything is working, then you can test the other models as well.

[464hee]

Hello, I am very interested in the TCGA training model. Can you share it? Thank you for your contribution @sophiajw