search

andreped / livermask

💥 Command line tool for automatic liver parenchyma and liver vessel segmentation in CT using a pretrained deep learning model
https://pypi.org/project/livermask/
MIT License
63 stars 11 forks source link

Segmentation is not performed - possible typo in Livermask.py? #10

Closed silgonel closed 3 years ago

silgonel commented 3 years ago

Hello Andre, and thank you very much for sharing your code! I have been trying to install and run your code on my PC, but unfortunately nothing happens after the first TQDM progress bar (no log information are shown, and even if everything seems to work fine, no segmentation is produced).

I have a doubt:

Line 73 of "livermask.py" if not curr.endswith(".ini"): continue

Could this be a typo (".ini" instead of ".nii")? Maybe my files are all skipped and that's why I don't get a segmentation nor a warning.

Many thanks! Silvia

andreped commented 3 years ago

Oh, yeah, that looks like something I would do :P Good catch, @silgonel !

I have fixed it now, in addition to some other minor things related to memory usage and whatnot.

Could you try reinstalling?

pip install git+https://github.com/andreped/livermask.git

If it does not overwrite the old version, try uninstalling before reinstalling.

Also, regarding prints, I would recommend using the flag --verbose if you want more information during deployment. See here for more information.

EDIT: I just now tried installing from the master branch myself, in a fresh virtual environment, and running livermask as a CLI. Seems to work, both for GPU and CPU inference, and single and batch mode, on my Windows laptop. Let me know how it works for you :)

andreped commented 3 years ago

I made it available as a release now, for other to try more easily. But if you still experience some issues, please, feel free to ask :]

Closing for now!

silgonel commented 3 years ago

Hello @andreped and many thanks for fixing it so quickly!! I managed to use your code on CTs and it really performs segmentation very well!

andreped commented 3 years ago

Great, @silgonel ! Happy to help. I am always trying to find ways to improve the tool, hence, if you find other mistakes or if there are features that you feel are missing, please, let me know :]