If I have a bunch of images that I could label and add to the training dataset but not the original videos, could I just skip the extract frames step and make the folder myself? I have already extracted frames from a set of videos; however, I would like to add more labeled images, and I do not have any more videos.
Yes! Just make sure your folder name is added to video_sets in the config.yaml (as if it came from an actual video; e.g., if your folder is newimages, add a dummy your_path/newimages.mp4). You should then be able to label the images normally, and most importantly, that folder won’t be ignored when creating the training dataset