This question is conceptual as I have a model working decently well and I am looking for different things to try out.
Research Problem: I am trying to classify each pixel in an image with values from 0->1. This is a regression problem.
Dataset: I have 30 images that are 1000x1000 and ground truth for those images.
Current Approach: For the training, I am extracting 65x65 patches to train a CNN which gives a label to the center pixel. Then in the evaluation of the model, I am striding through the full image labelling each pixel.
Is there any other methods I can try out.