I have 400 frames already labeled. I need to resize the labeled set to be significantly smaller (for faster training). However, I will lose the labels in the corresponding .csv and .h5 files, correct? Is there a way to resize them while keeping the labels in the .csv file?
create_training_dataset you can edit the
pose_cfg.yaml that is created and set the
global_scale (default is 0.8) to be smaller, i.e. 0.4, etc) that will decrease the frame size.
See box 2 in the Protocol paper, or here: https://github.com/AlexEMG/DeepLabCut/blob/master/docs/functionDetails.md#f-create-training-datasets