Self Driving Car

This project provides accurate prediction of steering angles of self driving cars. It was inspired by Udacity's Self driving car Nanodegree and NVIDIA's End to End Learning for Self-Driving Cars.

The code uses Convolutional Neural Networks are used to predict the streering angle corresponding to an input image of the road.

Requirements

To install the requirements, execute from your working directory, the following code in the terminal.

pip install requirements.txt

Dataset

The dataset can be downloaded here and should be extracted into the dataset folder at the root of the repository folder

Preprocessed files & train model

You can download the already preprocessed files and use them directly to train the model:

• labels
• preprocessed

If the pretrained model model_adam_mse.h5 is used, only the step 4 of the procedure bellow will be left to execute.

Procedure

1. Make sure the dataset is in the dataset folder.
2. Run load_preprocessed_data.py to get dataset from the folder and store it the preprocessed version in a pickle file.
3. Run train.py to load the preprocessed data from the previously savec pickle file and perform the training operations.
4. Run self_driving_car.py to test your results on the video.

References:

• Behavioral Cloning Project
• Mariusz Bojarski, Davide Del Testa, Daniel Dworakowski, Bernhard Firner, Beat Flepp, Prasoon Goyal, Lawrence D. Jackel, Mathew Monfort, Urs Muller, Jiakai Zhang, Xin Zhang, Jake Zhao, Karol Zieba. End to End Learning for Self-Driving Cars
• Inspired from this github repository: https://github.com/SullyChen/Autopilot-TensorFlow