A python script that finds the onscreen chessboard and then calculates the best move in the position using the locally downloaded stockfish and then plays it. It also creates the FEN(Forsyth-Edwards Notation) and a link to the lichess online stockfish engine for further analysis with gui. Useful for analyzing puzzles found in 2D images or even as a cheating tool on online chess games. Can also be used in combination with a trained chess AI in order to determine its elo ranking on popular chess platforms by having it play automatically online.
The keras model was trained for image classification using this dataset https://universe.roboflow.com/chess-project/2d-chessboard-and-chess-pieces by splitting the chessboards that were labeled for yolo in squares using the split.py script. Some extra images were also added.
Clone the repository locally by entering the following command:
git clone https://github.com/johneliades/chess_cv.git
Or by clicking on the green "Clone or download" button on top and then decompressing the zip.
Then install the missing libraries in a virtual environment:
Windows
python -m venv .venv && .venv\Scripts\activate && pip install -r requirements.txt && deactivate
Linux
python -m venv .venv && source .venv/bin/activate && pip install -r requirements.txt && deactivate
Go to this url https://stockfishchess.org/ and download the latest version of stockfish. The engine must be in the same folder named stockfish.exe.
Then you can run:
.venv\Scripts\activate (or "source .venv/bin/activate" for linux)
python chess_cv.py b
for black, or
.venv\Scripts\activate (or "source .venv/bin/activate" for linux)
python chess_cv.py w
for white.