Training of a convolutional neural network for autonomous vehicle Driving

In the world of electric vehicles, autonomous driving symbolizes the present and future where research is mainly focused. This paper shows the process to develop the training of an intelligent driving system based on artificial vision for an autonomous electric vehicle, making use of a convolutional neural network architecture, which are fed by a set of images of a route from three cameras, left, center and right, positioned in front of the vehicle, and the instantaneous direction of each third of images. The objective is to train a neural network to obtain a model that can autonomously make a decision about the angle that the vehicle should have in each input image frame, coming from a single camera mounted at the center of the vehicle and therefore that the vehicle covers the route autonomously. The images must be preprocessed to enrich the dataset, this is done in PYTHON specifically in Google Colab. In this first stage, the data set is obtained for preprocessing and performance testing of the model trained in the UDACITY autonomous driving simulator.