Stigmergy based mapping of multi agent systems

I want to extract the features from an image like shapes of an object, size of an object and coordinates of an object using any machine learning technique like artificial neural network. After this we have to create a model such that there are 2 agents and a medium like a router in between the two of them. And in which the first agent loads any image and then extracts the features of that image and then passes it to the medium like the router and the second agent is able to read the image from the medium. And displays that image with the results.