This is a fork of Learning to Act by Predicting the Future, by A. Dosovitskiy and V. Koltun.
This repo allows for DFP agents to take in additional vision inputs such as depth, semantic labels, flow, etc...
Please see the original repository for installation instructions.