I have been working with deeplabcut and am having trouble with getting the software to analyze my video. I went to dlcmodels/…/train and chose pose_cfg.yaml as the config file. This does not work and the error is “TypeError: ‘NoneType’ object is not subscriptable”. When I do the config.yaml file, the error is “IndexError: index -1 is out of bounds for axis 0 with size 0”. I honestly have no idea what is wrong at this point. How many iterations do you usually have to for analyze video to work? Thanks in advance!
config.yaml file is always the “config” file. The other “config” files, like
pose_cfg.yaml, contain information used in various steps of the DLC workflow.
When you trained the network, there was a setting for “save iterations”. That’s how many iterations have to run before saving a snapshot model. If you stopped training before that many iterations, then nothing was saved.
Depending on your video, labeled frames, batch size, and other factors, a few thousand iterations might get you some rough tracking to “test” out the system. For better analysis you’re looking for tens- to hundreds-of-thousands of iterations.