Individuals are mislabeled when analyzing the video #2877

Activity

1. alagithub

   assigned

   n-poulsen

   on Feb 12, 2025
2. n-poulsen commented on Feb 17, 2025

   

   n-poulsen

   on Feb 17, 2025

   Last edited by maximpavliv

   Contributor

   More actions

   @alagithub currently if the number of individuals you specify in the config.yaml is different to the number of tracks asked to be tracked (through the n_tracks parameter in analyze_videos), the names will be updated. So to keep the original names, as of now you'll need to update your project config.yaml before running analyze_videos.

   As this isn't ideal, I'll fix this issue so it is handled better when the number of individuals to track is smaller than the number of animals named in the config.yaml.
3. n-poulsen

   added

   feature request

   on Feb 17, 2025
4. n-poulsen

   closed this as completedin #2884on Feb 17, 2025
5. n-poulsen commented on Feb 17, 2025

   

   n-poulsen

   on Feb 17, 2025

   Contributor

   More actions

   @alagithub This has been fixed in #2884 - if you update DeepLabCut to the latest version on GitHub (2.3.11) the individual names should be set correctly as M1-M5 when using fewer than 5 mice. To upgrade your DeepLabCut version to the latest version, you can run in your DEEPLABCUT conda environment (as described in the docs):

   pip uninstall deeplabcut
   pip install --upgrade "git+https://github.com/deeplabcut/deeplabcut.git#egg=deeplabcut"

   You can also use the new animal_names parameter when calling analyze_videos with a greater number of individuals than defined in your project config.yaml, e.g. to analyze with 6 mice you could set:

   import deeplabcut
   
   deeplabcut.analyze_videos(config, videos, shuffle=1, animal_names=["M1", "M2", "M3", "M4", "M5", "M6"])

   Note that the individual names in the predictions file will only be those set in the project config.yaml if the number of mice your are setting when analyzing the videos is smaller or equal to the number of mice defined in the config.yaml (we'll take the first n labels when looking for n mice in the videos).

   If you want the labels to match the ones in your config.yaml file without updating the code, you'll need to edit the config.yaml before you run video analysis every time so it matches the number of mice you're looking for. Let me know if you have any other issues or questions!
6. alagithub commented on Feb 18, 2025

   

   alagithub

   on Feb 18, 2025

   Author

   More actions

   Thank you for the fix. When I am actually running the 5 Mice trained on trace 2 mice, its giving me a user warning UserWarning: No optimal solution found. Employing black magic…
   warnings.warn(“No optimal solution found. Employing black magic…”). The program is not moving forward after that. This doesn't happen when I am analyzing the video with 5 mice. Can you please let me know how to work around this.
7. n-poulsen commented on Feb 18, 2025

   

   n-poulsen

   on Feb 18, 2025

   Contributor

   More actions

   Is the quality of the predictions still good? This is a warning when tracklet stitching does not find a perfect solution; but if the tracks that are output are good you can ignore the warning