Extracted but not labeled frames

Hi !

I pretty new to the deeplabcut community, but I am wondering if during the labeling process the frames that are extracted but not labeled are treated as ''there is nothing to be tracked here" or if there are just not taken into account ? I am using it to track mice in an openfield.

Many thanks !

I think is better to delete them. There is a function for this, is
deeplabcut.dropimagesduetolackofannotation(config_path)

1 Like

Thanks :slight_smile: I felt like identifying ‘frames where there is nothing to track’ could however also be useful for training phase.

Not labeling in a frame is also good information for the network, so if you don’t see an object in a frame, don’t label it. But, if you have nothing absolutely in a frame it gets ignored, so you should consider having something labeled (if that makes sense).

Great, it makes sense ! Thank you for your time :slight_smile:

1 Like