A collection of projects for the self driving car nanodegree.
.python_code folders contain .py scripts that were used to develop the final solution
My main concern during the project was not the accuracy of the final model. That can be easily increased by using less max pool in layers and increasing the depth of the network at the cost , of course, of training time
I was more concerned in creating the tools to explore the problem and get an intuition of how the problem works and how the network design affects the final result.
In the 02.project.traffic.signs.python_code folder, all the code created for this project is contained.
The most important lesson from building this simple project was TAKE CARE OF YOUR DATA Making them as unbiased as possible and as immune to variability as possible. The model architecture was not so important as was the ease any of those model changed behavior if the data where not perfectly conditioned.