This repo contains a series of notebooks to help Metis students complete the DG/Metis project. The goal of this project is to train a segmentation neural net to recognize different kinds of terrain and land-use.
The setup instructions will help you install a few packages for geospatial data and deep learning. Some of these packages can be pretty hard to install, so you might still have some challenges.
Once you have geopandas, rasterio, tensorflow, and keras installed, you should be able to proceed to the numbered jupyter notebooks. These will take you through each step of acquiring satellite images and geospatial data, setting up a training set, and then training a neural net.
You can find more background about in my blog post on the DeepCore blog: http://deepcore.io/2017/06/06/Metis_Project.html