Self-driving-robotic-car-using-Artificial-Neural-Network

Supervised learning to train a robotic car
1-prepare-data : prepares data on which the NN is trained
2-Train-ANN : trains the 3-layered ANN on the data prepared using above code and saves the weight matrices
3-ANN-move : loads the weight matrices and makes predictions to move the robot along the path
The car is mounted with a camera-phone and can classify the images as one among forward, left, right, obstacle and zebra-crossing.
Sample images on which the robotic car was trained on:
The image taken by camera is first converted into 20X20 grayscale and fed into the network: (the image below doesnot represent network architecture; rather it is a screenshot from the webapp that streams the video as the robotic car travels along the path)