Determining Chess Game State From an Image

TL;DR

This paper introduces a large synthetic dataset and a novel end-to-end computer vision system that accurately recognizes chess positions from images, significantly outperforming previous methods and adapting quickly to new chess sets.

Contribution

It presents a new large synthetic dataset and a combined vision and deep learning system for chess recognition, with effective adaptation to unseen sets using few-shot learning.

Findings

Error rate of 0.23% per square, 28 times better than prior state-of-the-art.

Achieves 99.83% accuracy on new chess sets with minimal data.

System effectively localizes and classifies chess pieces from images.

Abstract

Identifying the configuration of chess pieces from an image of a chessboard is a problem in computer vision that has not yet been solved accurately. However, it is important for helping amateur chess players improve their games by facilitating automatic computer analysis without the overhead of manually entering the pieces. Current approaches are limited by the lack of large datasets and are not designed to adapt to unseen chess sets. This paper puts forth a new dataset synthesised from a 3D model that is an order of magnitude larger than existing ones. Trained on this dataset, a novel end-to-end chess recognition system is presented that combines traditional computer vision techniques with deep learning. It localises the chessboard using a RANSAC-based algorithm that computes a projective transformation of the board onto a regular grid. Using two convolutional neural networks, it then predicts an occupancy mask for the squares in the warped image and finally classifies the pieces. The described system achieves an error rate of 0.23% per square on the test set, 28 times better than the current state of the art. Further, a few-shot transfer learning approach is developed that is able to adapt the inference system to a previously unseen chess set using just two photos of the starting position, obtaining a per-square accuracy of 99.83% on images of that new chess set. The code, dataset, and trained models are made available online.