What I was busy with on New Year’s night and several nights after — integrating a physical chess board with a virtual one. Move a piece on the physical board – it moves on the virtual one, and the move is recorded. In the video — the first “public” test, a sort of “family concert”. For recognizing the board, I use a neural network based on resnet18, trained on 2300 images of pieces on a board. For working with images, I use OpenCV, PIL, Numpy.
The main task was simply to automatically record the games that Lisa and I play almost every day.
The prototype works, now the task is to throw everything away and start from scratch properly. Not with scotch tape, an easel, and a cane, and not with crappy code, so it wouldn’t be embarrassing to share. From simple but interesting plans: to remove the need to press space after each move, connect stockfish, in-depth game analysis on the fly, to produce a farting sound or the trademark ta-da-da-daaaam when one of us makes a stupid move. If anyone is interested in joining — give a shout.
Hi, I'm Rauf Aliev. 