Transferring labeled frames to new project

I am working with a well-trained network (~1700 frames labeled, trained over 12 iterations), but for organization/access reasons I need to move the network in order to update it with new frames. I saw a previous topic (Merging deeplabcut training sets) suggesting that I could just copy over the labeled-data folder into my new project folder to transfer the labeled frames to the training dataset.

However, when I run create_training_dataset, I get an error related to the ‘Project Scorer’, which I changed to my name in the new project folder. I changed all of the Collect_data filenames to my new project scorer name, but I realized there are references to the scorer in the files themselves, too.

Is there an easy way to adjust the project scorer name across all of the labeled-data files? Or do I need to just revert back to the original project scorer name for this to work?

Hi Victoria,

If the labeler in the .h5 files is not correct, you can run: deeplabcut.convertcsv2h5and change the scorer name: i.e.:

deeplabcut.convertcsv2h5(configpath, userfeedback=True, scorer='NewName')

1 Like