first, use 'get screenshot in a video in frame' to get all your picture, they maybe tens of thousands
second, use 'cut image' to get the part you want
finally, open where you save your cut image and classify them for further use
Valarzz/make-picture-data-set
When it comes to the image processing, deep learning and the neural network, we're usually lack of data set, this repository can help you get a huge data set easily.
Python