A new deep-learning algorithm for drug-discovery based on images, requires splitting a file containing ~3000 chemical compounds in png files containing individual 2D 200 x 200 p