Smart car Project "Mercury"

Traffic sign recognition

Mercury is a WIP prototype of a self driving car, the first model focuses on traffic sign recognition using Deep Learning (CNN).
Dataset : German Traffic Sign Dataset

Here's an overview of Mercury 1.0 :

Signs the AI currently recognizes :

NB : The AI is only trained with 5 signs (50, 120, Right, Left, Stop) but the variation the dataset allows it to recognize all the signs displayed above.

1. Dependencies

• Python3.7, Jupyter, NumPy, SciPy, TensorFlow, Keras, Matplotlib, Pandas, OpenCV, Skimage, pillow
• OS (RaspberryPi) : Raspbian

2. Hardware

• RaspberryPi 3 B+, 8MP PiCamera, 2DC motors, L298n h-bridge, 2 KY 033 optic sensors

NB : We used optic sensors to keep the robot movement in a straight line, if you are planing to use 4 DC motors or other alternatives you won't need them

Video demonstration

Data processing and model training

RaspberryPi comand