-
-
Notifications
You must be signed in to change notification settings - Fork 1.8k
Description
Is there an existing issue for this?
- I have searched the existing issues
Bug description
Hi, hope someone can help me figure out what's wrong because I have run out of ideas.
I'm running video analysis with maDLC, with 4 mice that are distinguishable (i.e., painted red, yellow, green and blue). They were labeled consistently in 400 frames, 8 bodyparts for each mouse. I've run the training for 75.000 iterations.
I analyzed short videos (few minutes each), and it assigns the bodyparts correctly, however it's like the mouse identity is randomly assigned. For example, it will identify the green mouse as mouseRED, the blue one as mouseYELLOW etc. This changes for each video that I try. Throughout the same video, the tracking performs very well and there are no switches, however it's like it didn't take into account the identity=True passed in the config file.
I made sure to pass the identity_only=True when doing the analysis of the videos. And I tried for different track methods, i.e., "ellipse" and "skeleton".
This is the screenshot from my config file:




Operating System
windows 11
DeepLabCut version
2.3.2
DeepLabCut mode
multi animal
Device type
google colab GPU
Steps To Reproduce
No response
Relevant log output
No response
Anything else?
No response
Code of Conduct
- I agree to follow this project's Code of Conduct