I have images which contains maybe 1000 circles of different sizes. Circles may overlap: it's not a simple picture.
I'd like to do one of two things: either (1) learn the circle size probability density function (binned), or (2) count the number of circles of each size.
Can anyone refer me to any references relating to such a regression from an image of a probability distribution? I'm interested in methods to constrain neural network output so the sum of values from the output layer is 1.0 (approach 1), or to count objects without localizing their position (approach 2).