The problem of fully autonomous cars is interesting but complex as a whole. The first step towards solving this problem is to get comfortable steering the car. We can do this by predicting the steering angle given a sequence of images or a video.
The dataset consists of 45,406 images from a camcorder in a real car ride. The data.txt file contains the steering angle for each image.
Credits: Sully Chen
This project is a keras implementation of DAVE-2 model from this Nvidia paper.
Model architecture
