Closed jumutc closed 2 years ago
@muellerdo I guess some guidelines for Multi-Path architecture usage are needed as well. As stated in the paper all paths should have coherent strides given in self.feature_maps
.
Hello @jumutc,
thanks for your great contributions to this repo, as always! :)
Novel architectures are always welcomed and congratulations to your publication.
I have switched the target to the dev branch and added a copyright header as well as a reference to your paper in the code.
@muellerdo I guess some guidelines for Multi-Path architecture usage are needed as well. As stated in the paper all paths should have coherent strides given in self.feature_maps.
If needed, could you add a note of this in the class doc string (between line 43 and 44)?
@muellerdo added clarifying comments to the docstring. Should I put something to Wiki?
@jumutc perfect!
Should I put something to Wiki?
Not needed. I will just copy & paste the class doc-string to the wiki. If you want to add more to it just let me know^^
@muellerdo all is set then! Thanks for a prompt review! Me and my team will contribute more as we keep on solving some challenging problems in the medical image segmentation.
Added documentation to wiki.
@jumutc, perfect and happy to hear! Good luck on further studies!
@muellerdo maybe Wiki should be corrected a bit as naming/description is incorrect: "The MultiRes variant of the popular U-Net architecture. It is an improved version of the standard U-Net and contains some small modifications to improve upon the already state-of-the-art U-Net model." I would change for: "The Multi-Path variant of the popular "plain" U-Net architecture. It is an improved version of the "plain" U-Net and contains a few modifications to improve upon the already state-of-the-art U-Net model."
Please consider the latest MDPI paper and the actual implementation.