Skip to content
master
Switch branches/tags
Code

Latest commit

 

Git stats

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
 
 
 
 
 
 
 
 
src
 
 
 
 
 
 
 
 
 
 

Behavioral Cloning for self-driving cars

Overview

In this project, we use a deep convolutional neural networks to help predict steering angles from camera images. To showcase this, we use simulator to collect data of good driving behavior and then use the images to train a model. A video with the footage of the car driving itself in autonomous mode is available in the repository.

alt text

You will find the code - using Keras - for this project is in the IPython Notebook. More details are available by reading the project notes.

The final model architecture :

Layer Description
Input 160x320x3 Image
Cropping 90x320x3 image
Normalization Normalization
Convolution 5x5 2x2 stride, valid padding, outputs 43x158x24
RELU activation
Convolution 5x5 2x2 stride, valid padding, outputs 20x77x36
RELU activation
Convolution 5x5 2x2 stride, valid padding, outputs 8x37x48
RELU activation
Convolution 3x3 1x1 stride, valid padding, outputs 6x35x64
RELU activation
Convolution 3x3 1x1 stride, valid padding, outputs 34x33x64
RELU activation
Flatten
Fully connected 8448 input, 100 output
Fully connected 100 input, 50 output
Fully connected 50 input, 10 output
Output 10 input, 1 output

About

Behavior-cloning for self driving cars (steering angle)

Resources

Releases

No releases published

Packages

No packages published

Languages