Converting single instance labels to multi-instance

[jackransomlovell]

Hi! We love SLEAP! Thank you all for building such a great piece of software.

I have ~500 frames of labeled data, with one instance. The skeleton used is mice_hc, plus one extra node for a cricket in the frame, where the cricket does not have any edges.

I trained a base_medium.single model, and despite the training dataset having a decent number of frames without the cricket in it, the model still wants to label the cricket in the frames when it is not present.

Any suggestions for how to solve this? More training data without the cricket?

See below for examples where the cricket is properly labeled and not, the network is labeling the rat's tail as the cricket in the wrong case.

Thanks again!

[jackransomlovell]

Changing peak_threshold and adding more training data seems to have helped. Closing.