Institutional Repository of Vilnius University
Šachmatų žaidimo sekimas realiu laiku naudojant neuroninius tinklus /
Abstract
dc:descriptionChess game tracking in real-time is an unsolved problem. Solving this problem would help professional and amateur chess players alike to track their chess games. This problem is usually solved by using expensive chess boards with digital sensors or writing down the moves made by hand. In this paper, an analysis of previous works is conducted. Many authors have tried to solve this problem, but only a handful of solutions are meant to track a chess game. The main issues of these works consist of the inability to automatically determine the orientation of the chessboard and the use of classifiers, which introduces extra steps when processing photos of the board. These extra steps produce inaccuracies, which lead to incorrect piece classification. A decision was made to use neural networks for object detection to solve these shortcomings. A new dataset was created by labeling chess pieces and their positions to train the neural network for chess piece detection. The trained model was used to develop a prototype system, which tracks chess games by automatically determining the chessboard orientation, detecting pieces and their locations, and identifying the move made. The proposed solution was evaluated by measuring its tracking accuracy and processing time and comparing it to existing solutions.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Rudys, Vytautas Petras,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD210647420&prefLang=en_US