Self-Driving Car Exploiting Supervised Learning via a CNN
Created in conjunction with Matteo Prata. This project was presented on the 11th of January 2019 in the Department of Computer Science of La Sapienza University in Rome, Italy. It is a Self-Driving Car 3D model built in Unity, with the task of line following. The model uses a Convolutional Neural Network to process images coming from three cameras put on top of the car and to perform predictions according to the current situation. Predicted values are the steering angle of the car. For any information about the system, look at the slides in the directory CV-pres.pdf, feel free to contact us for additional details.
In order to try out our CNN model, please download our 3D Unity worlds and neural model at this page. Locate either the model or the worlds in the same directory as the Main.py file. Run Main.py file and start the game application (choose ScreenCapture mode). Be sure to have imported all the dependencies required by the project (OpenCv, Keras...).