CNN and 3DCNN reference:https://github.com/rysmarie/MotionRecognition/tree/master/

This code requires UCF-101 dataset. All of video files need to be put in a file.

Options of 3dcnn.py are as following:
--batch batch size, default is 128
--epoch the number of epochs, default is 100
--videos a name of directory where dataset is stored, default is UCF101
--nclass the number of classes you want to use, default is 101
--output a directory where the results described above will be saved
--color use RGB image or grayscale image, default is False
--skip get frames at interval or contenuously, default is True
--depth the number of frames to use, default is 10