I have an image (attached) bearing diseased crops. I have 300 such images. I need to extract the diseased portion and create a bounding box around it. But as the colour change of diseased part is only slightly different, I am not able to obtain the diseased regions effectively. Currently I am doing neural network training but am open to any other algorithm suggestion to obtain diseased areas from 300 input images.
The image with red bounding boxes is what the output should look like. DESIRED OUTPUT: