AFFORDABLE VISION-BASED SYSTEMS FOR REAL-TIME CHESSBOARD DIGITIZATION

The automatic recognition of chessboard states has significant applications in various domains, from enhancing online chess platforms and educational tools to enabling robotic interaction. While high-performance vision systems and complex robotic setups can achieve this, their cost and complexity often limit widespread adoption. This paper explores the feasibility and methodology for developing affordable, vision-based systems for real-time chessboard digitization. We leverage advancements in deep learning, particularly lightweight Convolutional Neural Networks (CNNs), combined with accessible embedded platforms. The proposed approach integrates image acquisition, chessboard localization, and individual chess piece recognition, culminating in a standardized digital representation of the board state. Our findings demonstrate that acceptable levels of accuracy and real-time performance can be achieved on low-cost hardware, making automatic chess digitization more accessible for a broader range of applications.