Research which tech stack and which tools to use for the computer vision

[yasmineben510]

• [x] what are the possible libraries algorithms?
• [x] Where can you run them? => Computer, not on arduino
• [x] can you already take pictures of something that could be a possible game figure? Do some tests with software?
• [x] Research solution for tracking the position of new pieces of the board

Research:

-> AruCo Marker :

• https://docs.opencv.org/4.x/d5/dae/tutorial_aruco_detection.html (cpp)
• https://mecaruco2.readthedocs.io/en/latest/notebooks_rst/Aruco/aruco_basics.html (python) -> Color detection Algorithm with openCV only:
• https://www.youtube.com/watch?v=aFNDh5k3SjU
• https://github.com/computervisioneng/color-detection-opencv

-> Computing distance of object from camera:

• https://www.geeksforgeeks.org/realtime-distance-estimation-using-opencv-python/

-> Using a pre-trained object detection model:

• https://www.youtube.com/watch?v=mNjXEybFn98

-> Detecting the chessboard:

• https://blog.roboflow.com/automated-chess-game-recording-computer-vision/

Notes:

• AruCo markers: very reliable, even with busy background/obscured areas ✔️, pieces need a flat head & camera abovehead ✖️
• Color detection: fast, real-time detection✔️, needs a clean background✖️
• Trained recognition model: clean chessboard without markers ✔️, long training curve, can be weak to environment ✖️

[yasmineben510]

@Svannia helps