Skip to content

Getting started with ADAPT

A good entry point into exploring the ADAPT processing software without requiring a flyable ADAPT payload is provided by the tutorial: ADAPT Processing of AirSim Data. This tutorial uses Microsoft AirSim as a drone sensor emulator running alongside ADAPT processes within a fully Dockerized environment. Camera images along with associated ground-truth depth and semantic segmentation images are rendered and saved to disk as you navigate a simulated drone with your keyboard through a forested mountain landscape. The pairing of raw imagery with ground truth allows training a semantic segmentation deep-neural-network model using ADAPT training tools. Finally, the trained model is deployed to process AirSim imagery in real time.