Deep Learning-Driven Object Detection of Chess Pieces for Precise Robotic Moves and Game Notation

Abstract

The paper explores the utilization of YOLOv8, a resilient architecture for detecting objects, in the field of transfer learning for accurately recognizing chess pieces. The objective is to streamline the execution of robotic movements and the recording of game moves within the framework of chess. By utilizing this technique, the study achieved outstanding performance measures, specifically a mean Average Precision (mAP) of 98.5%, precision of 98.2%, and recall of 97.9%. The assessment consisted of a thorough evaluation that included a wide range of testing situations. The results demonstrated that the model consistently and accurately detected chessboard pieces with detection rates ranging from 89% to 94%. The model demonstrated its versatility and dependability in accurately identifying chess pieces, regardless of their placement or density on the board. The results emphasize the model's resilience and its capacity for practical use in robotic manipulation and game notation specifically in the context of chess. This study establishes the foundation for future progress in AI-powered gaming analysis and opens up possibilities for the wider use of this technology in various AI applications. The model's ability to reliably recognize chess pieces is a significant advancement in the field of robotics and intelligent systems. This achievement has the potential for very precise object identification and manipulation and paves the way for additional research and improvement in real-time applications.