lclamotte / chesscv

Converts 3D chess positions to 2D representations

chesscv

Converts 3D chess positions to 2D representations

References

ImageAI custom object detection tutorial: https://imageai.readthedocs.io/en/latest/customdetection/index.html Chessbase India: https://www.youtube.com/channel/UCIsEhwBMPkRHsEgqYAPQHsA/videos

Questions/Notes:

• Should test data have 64 annotations for each square or <= 32 for pieces?

• If each square: -- An empty square could be represented with 0a1 (no piece on a1), 0f7 (no piece on f7) -- A white square could be wbg6 (white bishop on g6), wpc7 (white pawn c7) -- Black square would similarly be bbb3 (black bishop b3), bkb8 (black king b8), bnc6 (black knight c6) -- Then each square has an annotation of the form [team] [piece] [square] -- Could use delimiters to parse or just switch logic based on first character

• Eventually feed into Stockfish for live advantage graph?