I want to train with my own dataset, but I want to train with image data instead of video data, how can I do that? Thank you.