kinetics-i3d
kinetics-i3d copied to clipboard
Shape of *.npy file?
In the sample code,the example video has been preprocessed, with RGB and Flow NumPy arrays provided. I want to test my own video, so I consider it might be a way to generate my own Numpy arrays and replace the example ones. For RGB, The provided *.npy file has shape (1, num_frames, 224, 224, 3). It seems that 'num_frames' means number of frames, '224,224' means heights and widths, ‘3’ means channel(RGB). I'm coufused about ‘1’, what does this mean?And its value?
By the way, what's the equation of norm of the logits tensor?
1 means batch_size