Martı Tag — maze-solving robot
Take a photo of a maze, and a small robot car drives through it.
- A segmentation model trained on about 17,000 maze images finds the open paths in the photo.
- The path map is cleaned into a thin skeleton; the user picks start and end points.
- A search algorithm finds the route, which is simplified to corner points and turned into commands relative to the car (forward, turn left, turn right), then optimised.
- A desktop app shows each step, animates the plan, and sends the commands to the Raspberry Pi on the car over a network connection.
Tools
PythonDeep learningImage processingPath planningRaspberry PiSockets
Photos
