Imitation Learning

This project aims to build a model for self driving cars based on immitation learning using TensorFlow 2. A new set of data was collected from CARLA driving simulator. The neural network and procedure is based on a paper about conditional immitation learning by Codevilla et al. (2018).

Files

• general_functions: various functions for reading files and visualization
• image_augmentation: the functions for generating augmented images
• network: the functions for constructing the convolutional neural network
• main: the main file to execute the process
• model_evaluation: reading test data and evaluating the model

Reference

Codevilla, F., Müller, M., López, A., Koltun, V. and Dosovitskiy, A. 2018 End-to-end Driving via Conditional Imitation Learning, IEEE International Conference on Robotics and Automation (ICRA).