Change Detection ================ Segmentation ------------ First step is to train model to segment the imagery to detect roads, and then find places where we detect roads in the new imagery but not in the old imagery. * Copy discoverlib from https://github.com/mitroadmaps/roadtracer/ (in lib/ folder) and put in segment/discoverlib. * Run segment/run.py to train the model. You may need to edit the paths at top of file. * Run segment/apply.py to obtain segmentation outputs. Again may need to edit some paths. * Run cmp-negative-road.go to obtain the detection where we find road in new image but not in old image.