Building a Traffic Sign Recognition Classifier (Self-Driving Car Engineer Nanodegree)

This repository contains the code in a form of a Jupyter notebook that could be used in trafic signs classification task. The preprocessed data is also presented in this repository in traffic-sign-data. The data contains German traffic sign images of 43 kinds (description). The model of convolutional neural network that performs classification can be found in Traffic_Sign_Classifier.ipynb. Please, follow the instructions below to prepare an environment to run this notebook. There is also Writeup.md where the pros and cons of the implementation described.

Preparing Environment

In order to be able to use traffic sign data in this repository, Git Large File Storage extension should be installed on the system. Installation instructions can be found on git-lfs.github.com.

To run Jupyter notebook, start a docker container with:

docker run --interactive --tty --rm --publish 8888:8888 --volume $PWD:/src udacity/carnd-term1-starter-kit:latest

Note, that the current directory will become your working directory in Jupyter notebook.

Then, copy a link from the console to your browser and start exploring the source code. The link will be similar but not equal to http://localhost:8888/?token=eb26e4a2b935c384dc3e0230a8181984f07da6be9df0c1b8.

Notice

Some functions from Traffic_Sign_Classifier.ipynb are provided by Udacity.com.