The goal of the project is to train a Deep Neural Network to replicate the human steering behavior while driving, thus being able to drive autonomously on a simulator. The network takes as input the frame of the frontal camera ( a roof-mounted camera) and will predict the steering direction at each instant.
- socketio
- eventlet
- numpy
- pandas
- flask
- keras
- Tensorflow
- base64
- io
- openCV
- PIL
- Python 3
- imgaug
For the purpose of simulation and training the neural network, i have used Udacity simulator.
To run this, start simulator in autonomous mode, and run on terminal/cmd python drive_car.py
The dataset consists of images (captured from the simulator) and steering angle. [
- crop the image
- Convert RGB to YUV
- Gaussian Blur
- Resize the image
- Normalize the image
- After selecting the final set of frames, we augment the data by adding artificial shifts and rotations to teach the network how to recover from a poor position or orientation.
- Data Augmentation techiques : Zoom, Width shift, Heightshift, Image brightness and Flipped image.
- All of these are implemented in order to diversify our dataset so that car moves more acuurately and smoothly.