chessqueries
chess
computer-vision

ChessQueries model weights

Resources: Demo · Paper [arXiv] · Code [GitHub] · SLCC dataset annotations [HuggingFace]

Try it online. The website uses a ChessQueries model further trained on proprietary data.

ChessQueries predictions across four chessboard datasets

ChessQueries can handle four datasets with various challenges.

Release weights for ChessQueries, a real-life chessboard recognizer jointly trained on ChessReD, ChessCog, and SLCC.

Files

Paper model Weights Size Manifest
ViT-L FP16 safetensors 743.35 MB JSON
ViT-B FP16 safetensors 248.95 MB JSON
ViT-S FP16 safetensors 63.15 MB JSON

Each manifest records the immutable download, SHA-256, preprocessing and validation references. Newer Lite weights are available separately at ChessQ Lite.

The original ViT-L FP32 weights, legacy ONNX/Core ML manifests and training checkpoint remain available for existing users.

Quick start

With Python 3.12:

python -m pip install "git+https://github.com/JSeytre/chessqueries.git@f5aaaf38212cc7839b03ccf383455053905b3f7e"
chessqueries-predict photo.jpg

The command downloads and verifies the paper ViT-L weights. CUDA uses FP16; CPU uses the same download with FP32 arithmetic.

For ViT-B or ViT-S, download the weights above, check the SHA-256 in their manifest, then pass the file explicitly:

chessqueries-predict photo.jpg --checkpoint /path/to/model.safetensors

Setup, evaluation and reproduction instructions are in the code repository.

Citation

@misc{seytre2026chessqueries,
  title         = {ChessQueries: Toward Better Chess Board Recognition},
  author        = {Seytre, Jo{\"e}l},
  year          = {2026},
  eprint        = {2608.30762},
  archivePrefix = {arXiv},
  primaryClass  = {cs.CV},
  doi           = {10.48550/arXiv.2608.30762},
  url           = {https://arxiv.org/abs/2608.30762}
}

License

The ChessQueries model weights are licensed under the PolyForm Noncommercial License 1.0.0. They may be used, modified and redistributed for permitted noncommercial purposes; commercial use requires separate permission. Third-party components retain their upstream terms; see the notices.

Contact

For questions about licensing, reach out to 'contact "at" chessq.org'.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Paper for joelseytre/chessqueries