threedle / iseg Goto Github PK

Hi,

I have been working on refactoring the code in this repository a bit for my purposes to run without any visualization built into it. I am able to run the test_decoder function without updating the weights with a subsequent click. However, I am running into a minor issue:

The resulting probabilities falling out of the mlp application here:

prob_tensor = mlp(input_tensor)

have the shape (1, 11595, 2), whereas the example hammer mesh has 69384 vertices and 23128 faces. Shouldn't the size of the prediction tensor correspond to the size of the mesh itself (probably the vertices, I suppose)?

For completeness, here's the link to the refactored files. The relevant par would be this Jupyter notebook.

Thanks in advance!

Dear iSeg authors,

I just stumbled over this repository and it looks like a super-cool tool. I work primarily in a biological-context, for which napari is the working-horse when it comes to visualization, rendering and interaction (see, for instance, this repo). I was thinking about turning your tool into an interactive plugin for napari (which is relatively easy, code-wise), but I did not want to start working on anything before ing the authors - you - for your consent, as there is currently no license on this repository.

If you were to add a permissive license such as BSD-3, I could freely use your code while of course adding a copy of your license to a to-be-created repository.

Let me know what you think :)