daHorse is a Python program that takes an image of a chessboard as input and converts its position to FEN (Forsyth-Edwards Notation) format. It uses Convolutional Neural Networks (CNNs) for chessboard recognition and the python-chess library for FEN conversion.
- Preprocesses the input image to find the chessboard contours.
- Extracts individual chessboard squares.
- Uses a trained CNN model to interpret the chessboard position.
- Converts the chessboard position to FEN format.
- Generates an SVG representation of the chessboard.
- Python 3.6 or higher
- OpenCV (cv2) library
- tensorflow library
- python-chess library
- Flask (for the optional web interface)
-
Clone the repository: git clone https://github.com/AliSinaDevelo/daHorse.git cd daHorse
-
Install the required libraries: pip install -r requirements.txt
To convert a chessboard image to FEN format, run the main script:
python src/main.py path/to/chessboard_image.jpg
The FEN position will be saved in the data/fen_output.txt file, and the SVG representation of the chessboard will be saved in data/chessboard.svg.
To use the web interface (optional), run the Flask app:
cd app python app.py
Visit http://127.0.0.1:5000/ in your web browser, upload a chessboard image, and get the FEN position and SVG representation.
The labeled dataset used for training the CNN model should be organized as follows: data/labeled_dataset/ |-- empty/ | |-- square_0.jpg | |-- square_1.jpg | |-- ... |-- pawn/ | |-- square_2.jpg | |-- square_3.jpg | |-- ... |-- ...
If you want to train your own CNN model for chessboard recognition:
- Prepare the labeled dataset in the above format.
- Run the training script:
python src/train_cnn_model.py
The trained model will be saved in the data/trained_models/ directory.
To test the entire "daHorse" program: python tests/test_main.py
To run unit tests: python tests/test_unit.py
Contributions are welcome! If you find any bugs or have ideas for improvements, please open an issue or submit a pull request.
This project is licensed under the MIT License - see the LICENSE file for details.