Format of input image training data

PreprocessTrainingData.m expects a directory of 8-bit grayscale png files containing the images and another directory of 8-bit grayscale png files containing the labels as image files. These images should 1024x1024 in size, but in theory could be smaller.

The label image files should have intensity of 1 to denote a feature that should be segmented and a value of 0 for areas that should not be segmented.

An example of this can be found in CDeep3M source tree

Example images:

mito_testsample/training/images

Example corresponding labels:

mito_testsample/training/labels