-
Notifications
You must be signed in to change notification settings - Fork 27
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
patchcommander rework #69
patchcommander rework #69
Conversation
…th rows and cols) because in that edgecase this patch would be completely empty
Pull Request Test Coverage Report for Build 11668319171Warning: This coverage report may be inaccurate.This pull request's base commit is no longer the HEAD commit of its target branch. This means it includes changes from outside the original pull request, including, potentially, unrelated coverage changes.
Details
💛 - Coveralls |
@martvanrijthoven had to fix one issue when the first patch was exactly one patch size outside the 0,0 coordinate Everything works now, please check the notebook if you want to see how it behaves. I have many plots there that show the output grid, also do some stress testing there and again: |
Heey Joey, Thanks for your work. I will try to look at it asap. But I already have a question. Would it make sense to give all arguments already as scaled? The patchsize, for example, is also already scaled. |
Hey Mart, I think we need to provide the settings in a way that makes sense from an ease of use perspective. When setting everything up the iterator:
I think scaling the offset internally makes most sense and is consistent with the rest. But I dont want to break older code, therefore for now I left it with a comment |
@martvanrijthoven, altogether nothing changed conceptually, but sampling patches around the edges with center=True or overlap is corrected now. All old code will work exactly the same, but cases where tissue lies right next to the borders of the images are now sampled correctly. If you want you can double check the notebook. Scaling/not scaling is another thing that can be changed, but right now I left it as it was so it can be integrated directly without any negative impact. If you agree with the merge ill delete the notebook, and we can integrate the PR. |
Complete rework of the patch commander, accompanied by a (temporarily) notebook that showcases that it works and how it works. Notebook is actually also a useful tool for setting up your patch config, doesnt support spacing setup yet though.
One remark. In the previous implementation the offset had to be provided in a already scaled form, while the other attributes do not. I remember that this was really had to debug when I was setting up the sampling logic for nnUNet. I think this should be changed such that it is scaled internally in the patch commander.