search

DeepLabCut / napari-deeplabcut

a napari plugin for labeling and refining keypoint data within DeepLabCut projects
GNU Lesser General Public License v3.0
52 stars 22 forks source link

Napari not responding when asked to label frames in deeplabcut #71

Closed bjmoraes closed 1 year ago

bjmoraes commented 1 year ago

Hi, I'm having trouble getting Napari to label frames in deeplabcut.

Everything goes well up to

deeplabcut.label_frames('/Users/bjm/Desktop/videos_cr/chrv12-7/test-BJM-2023-05-19/config.yaml')

Napari opens but the rainbow wheel keeps turning with nothing happening and the icon reports "Application Not Responding", so I am not even able to select the frame files to label.

I'm on a MacBook Pro, 13-inch M1, 2020 running macOS Ventura 13.3.1(a). I installed deeplabcut according to developer instructions without issues. Python is up to date:

Python 3.9.16 | packaged by conda-forge | (main, Feb  1 2023, 21:38:11) 
[Clang 14.0.6 ] on darwin

I also re-installed Napari for deeplabcut following its website instructions.

Help would be appreciated. Cheers, BM

bjmoraes commented 1 year ago

I will close this since I haven't had any answer for 20 days and I have worked around the issue without solving it.

MMathisLab commented 1 year ago

Hi @bjmoraes I can't reproduce the issue you describe and can label. I recommend to launch the GUI from the terminal vs and IDE.

bjmoraes commented 1 year ago

Hi @MMathisLab, thanks. That was the workaround I followed. Using the GUI all works fine; only when trying to label using Terminal commands do I get stalled. In the meantime, I have another issue using the Mac, iteration loss numbers are huge (no issue using a PC with the same data) but I may discuss this on a different thread.

MMathisLab commented 1 year ago

Hey! I just hit this Mac tensorflow metal issue on one dataset; you can change (lower!) the learning rate -- namely before starting to train and after creating training dataset go into the dlc-models, .. train folder, open pose_cfg.yaml and drop learning rate. @n-poulsen is working on automatically adjusting this for MacOS users now (I think ;), but that was the solution for me and him!

bjmoraes commented 1 year ago

Thank you! I will try that out and give you feedback on it :)