This game allows users to play against an AI opponent that makes moves based on a minimax algorithm with alpha-beta pruning and a pre-trained convolutional neural network (CNN) model.
- Clone this repository to your local machine.
- Install dependencies using pip (tested using Python 3.11.5): pip install -r requirements.txt
- Ensure you have a working version of Stockfish installed on your machine. You may need to modify the path to Stockfish in the code if it's installed in a different location.
To run the chess game, execute the ChessCNN.py file. The GUI window will open, allowing you to play against the AI. Enter your move in UCI notation (e.g., e2e4) and click "Make Move" to confirm your move. The AI will then calculate its move and display it on the board. If you would like to create your own dataset and model, remove the chess_model.h5 and dataset.npz files. Then, replace the number in the main method inside of populate_dataset(1000) with the desired amount of board datasets you would like to use.
- GUI: User-friendly interface built using tkinter, making it easy to interact with chessboard and make moves.
- AI Opponent: Challenge yourself against an AI opponent that utilizes both traditional algorithms like minimax and neural network-based evaluations.
- Chess Engine Integration: Incorporates Stockfish, a powerful open-source chess engine, for evaluating board positions and generating optimal moves.
- Convolutional Neural Network: Train and use a CNN model to predict board evaluations, enhancing the AI's decision-making process.
- Dataset Generation: Automatically generate a dataset of chess board positions and evaluations for training the neural network model.
Click on the thumbnail above to watch the video demonstration of using this program.