Is there an existing issue for this?

• I have searched the existing issues

Bug description

I think it's the scale issue: when the mouse is too large compared to training data, superanimal_mouse_topview model might fail to extract pose, returing a empy list rather than a dictionary with "coordinates" and "confidence" as keys. Then _average_multiple_scale_preds function fails to query information.

I add the three lines before line 90 to skip such situations:

if not isinstance(pred,dict):
    xyp[scale_id] = np.nan
    continue

Besides, would it be a good idea to give the statistics of mouse size in training data, so that we can roughly estimate what values might be suitable for scale_lsit to __init__ SpatiotemporalAdaptation instance.

Operating System

Ubuntu

DeepLabCut version

2.3.3

DeepLabCut mode

single animal

Device type

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