This repository contains code for reproducing the experiments described in our master's thesis. See README in each folder for usage instructions.
Short overview of the folders:
common: Contains scripts and Dockerfiles shared by multiple experiments.datagen-*: Used to generate datasets for the training.train-*: Used to train the models.models: Contains the trained TransFuser and InterFuser models.ex-[transfuser|interfuser]-[0.9.10|0.9.14]: Used to evaluate trained models in CARLA benchmarks.ex-kia: Used to evaluate trained models on real-world data.ex-transfuser-leaderboard: Used to submit TransFuser models for evaluation on the CARLA Leaderboard.