Open carloalbertobarbano opened 2 years ago
I can make an attempt. Will start with numpy on a new branch.
I have made an implementation with numpy backend in this PR: https://github.com/EIDOSLAB/torchstain/pull/28
However, it depends on spams
, and hence, some work likely still remains.
If spams
will be added, implementing the tf and pytorch backends should be trivial, as the main challenges rely on the two tasks that spams
solves.
For more information, see the aforementioned PR.
Add support for Vahadane normalization, as in https://github.com/Peter554/StainTools