imagine we have a stream of pixels coming in one by one. They form a video, but we don't know the width and height of each frame. Here we try a bunch of guesses for the dimensions, before locking in on the correct values
79
290
3K
0
256
Download Video
@matthen2 Very interesting! A few wild guesses: - minimize time derivative on small kernels (eg L2 or L1 norm on pixelwise diff on kernels - eg 3x3, 5x5 - of successive frames) - minimize spatial derivative (eg sobel) - use a img classification CNN :) No idea if will work. Just a hunch.