Chess & Computer Vision for Raspberry Pi

This is my first project for Raspberry Pi and since I am also interested in chess I wanted to make a system able to track a chess game and show the evaluation and best moves in real time.

My equipment:

• Raspberry Pi 4 model B 8GB RAM
• Raspberry Pi Camera Module 3
• Mini portable chess set
• Red/Green acrylic paint

I decided to paint my chess pieces red/green to enhance their visibility and ease of detection by the computer vision system against the background.

The chessboard is detected with the function cv2.findChessboardCornersSB(), then I compute the perspective transform with cv2.findHomography()

Chess pieces are detected with a YOLOv5n model with runs at 0.6 FPS on the Raspberry Pi. The model is trained on this publicly available dataset: Raspberry Turk Chess Dataset

Model weights are provided.