redundant patches in my training data

[chetanreddym]

Hello Marvin, Thanks for sharing your code.
I am trying to train the network on my data to do a binary segmentation of my input images.
Now i am using the crop_patch option to crop my input images into 256x256 patches, but the problem here is while doing so, i am ending up with few patches which solely consists of only one class(for example only background).
How can i eliminate such patches and use only the ones which contain both the classes of my interest.