Autonomous Driving using Deep Learning and Behavioural Cloning

This repository is to support the code for this blog. It contains the code to train a Deep Neural Network to drive a car!

Introduction

There are 2 phases for this:

1. Training
2. Testing

Training

The train.py file provides the code to train the Neural Network.

usage: train.py [-h] [--data_dir DATA_DIR]
                [--steering_correction STEERING_CORRECTION] [--epochs EPOCHS]
                [--batch_size BATCH_SIZE] [--model_path MODEL_PATH]
                
  --data_dir DATA_DIR   Path to Image directory and Driving log CSV file
  --steering_correction STEERING_CORRECTION
                        Steering Correction to applied to left and right
                        camera images
  --epochs EPOCHS       Number of Epochs to train the model+
  --batch_size BATCH_SIZE
                        Batch Size for training.
  --model_path MODEL_PATH
                        Path to save the trained model.

Testing

The drive.py file provides the code to test the trained Neural Network.

usage: drive.py [-h] model [image_folder]

  model         Path to model h5 file. Model should be on the same path.
  image_folder  Path to image folder. This is where the images from the run
                will be saved.

Credits

The drive.py is based on Udacity CarND Behavioral Cloning Project which can be found here