SLEAP training stopped predicting frames

[rubytseng18]

Hi,

I have been running into issues while training my model. It would train for both the centroid and centered instance, but then re-run the centered instance. Then it would say that it predicted on 0 frames. I haven't had this issue before, and the only difference I made was labeling more frames.

These are the details regarding the behaviour set up and model:

Behaviour set up: 1 single black mouse with a Neuralpixel scope. The wire is also in frame.

Model: a multi-animal top-down

The most recent successfully run model statistics:

I decided to add more frames and ever since then it wouldn't let run and would encounter the problem where it would run and predict on 0 frames. I tried only predicting without training and it would still not work.

The error code from the miniforge prompt is attached here:

The error code from miniforge prompt after training using single mouse here:

I would appreciate any tips and advice on how to continue to train and improve the model. Please let me know if you need more information from me.

Thank you so much!

[eberrigan]

Hello @rubytseng18,

It sounds like the centroid model training might have failed to find your animal. Can you also show screenshots of the receptive field size? We have some example notebooks to help with that -- https://github.com/talmolab/sleap-nn/tree/main/example_notebooks.

Thanks!

Elizabeth

[rpainaik]

Hi! I'm having issues with this as well. Were you able to find a solution?