/ai_cnn

Design and implementation of a convolutional neural network, to control the steering of a model car given images of a simulated road

Primary LanguagePython

ai_cnn

This is the design and implementation of a neural network able to drive a car autonomously.

How to build

In order to build this CNN the following additional packages are required:

  • tensorflow (with the optional gpu support recommended)
  • opencv
  • matplotlib
  • keras

All of them available within the conda packages list

suggested set-up

use the following commands to create a virtual environment using Miniconda3:

conda create -n tensorflow_env python=3.5
source activate tensorflow_env
conda install tensorflow-gpu opencv matplotlib pydot keras

How to train

To train this CNN you need to go to the folder where your images are... and execute the python-script called to_npz.py, in order to compress the images into an .npz compressed format, and the move this .npz file into the folder called dataset/<the_name_of_your_dataset> in the root of this project (where this README.md is).

The execute the following commands:

source activate tensorflow_env
python3 interface.py --dataset=<the_name_of_your_dataset>