ChessAI - Chinese Chess Game Analyzer

ChessAI is a groundbreaking tool that brings together computer vision, chess algorithms, and advanced analytics to revolutionize the Chinese Chess analytics landscape. With ChessAI, you don't need expensive electronic boards to analyze your games. Simply use your regular board, set up a camera to capture the position, and let ChessAI do the rest.

• Main source code: chesssai.
• Deep Learning / Data Preparation: dnn_models/data_preparation - Currenly only support for Chinese Chess (XiangQi), contact me for the license and the source code of the data preparation tool.
• Deep Learning / Training: dnn_models/training.

Roadmap

• Chess position detection.
• Chess engine integration.
• Move suggestion.
• Deep learning model for chess board detection (No need to use ARUCO markers).

Environment setup

• Requirements: Python 3.9, Conda, Node.js 18+.
• Clone this repository.

git clone https://github.com/vietanhdev/chessai --recursive

• Create a new conda environment and the required packages.

conda create -n chessai python=3.9
conda activate chessai
pip install -e .

• Install Node.js packages and build the frontend.

cd chessai/frontend
npm install
cd ..
bash build_frontend.sh

Build chess engine

• This project uses godogpaw as the chess engine.
• Install Go.
• Build the engine.

cd godogpaw
go build

• Copy the executable file (godogpaw*) to the ./data/engines folder.

Run the app

ENGINE_PATH="data/engines/godogpaw-macos-arm" python -m chessai.app --run_app

Replace ENGINE_PATH with the path to the chess engine executable file.

Data preparation & Training

This project uses computer vision and deep learning to detect chess pieces and chess board position.

AI flow for chess position detection:

• Go to dnn_models folder and follow the instructions in the README.md file to prepare the data and train the model.
• NOTE: Only training source code and pretrained models are included in this repository. The data preparation scripts and the training datset are not included. Contact me for the license and the data.

References

• This project was initially built for Hackster's OpenCV AI Competition 2023. Hackster Project: ChessAI - Chinese Chess Game Analyzer.
• Object detection model (for chess pieces) is based on YOLOX - License: Apache 2.0.
• Chess engine: godogpaw - License: MIT.
• UI components: shadcn-ui - License: MIT.