For this project we have a set of satellite images acquired from GoogleMaps. We also have ground-truth images where each pixel is labeled as road or background.
Our goal is to train a classifier to segment roads in these images, i.e.
assigns a label road=1, background=0 to each pixel.
Evaluation Metric: F1 score