Python/AI

Vehcile Detection
This project uses an SVM (Support Vector Machine) in order to detect vehicles in a video.

The SVM was trained on a dataset of 30,000 images, acquired from the KITTI database, GTI Database, and the Udacity Annotated Dataset. Please visit the Project Github for more information.

- KITTI Project
- GTI Project
- Udacity Database

Lane Finding

This script uses Python3 and OpenCV2 to detect lane lines on the road. The script then paints a polygon with a green center around the detected lane.

Lane Curvature, and estimated distance from left and right lane line are also calculated.

Please see the Project Github for more information.

Behavioral Cloning

This project uses a Convolution Neural Network trained with driving behavior. "Good behavior" was captured from a simulator (by means of a video recording), and used to train the network. The network was then implemented as a guidance system, using video images as an input, and steering direction as an output.

This project took advantage of the neural architecture proposed by Nvida in this paper.

Please see the Project Github for more information.

Image Classification
This project uses Tensor Flow in order to train a CNN (Convolutional Neural Network) in order to classify traffic signs in an image.

Tool : Tensor Flow
Unique Images : 34,000
Modified Image Database : 200,000
Model Accuracy : 85%

Please visit the Project Github for more information.