nightmare-nectarine/segmentation-based-imitation-learning-in-CARLA

This repository provides autonomous driving imitation learning agents trained in CARLA simulator for route navigation tasks. The model takes a front-view RGB camera image, a segmentation image, traffic light status, current speed measurement, and a high-level command (GoStraight, TurnLeft, TurnRight, FollowLane), and outputs vehicle driving commands (throttle/brake, steer).

The models are trained on this dataset. For example usage, check out this project.