29 January 2025
09:00 Master's Defense Room 85 of IC 2
Theme
Chessboard Digitization and Human Move Prediction through Machine Learning
Student
Heigon Alafaire Soldera Pires
Advisor / Teacher
Helio Pedrini
Brief summary
Chess is a strategic game that has been widely studied due to its complexity and the cognitive skills it requires. The analysis of chess games has been improved by the use of machine learning techniques, which allow not only scanning chessboards to identify the positions of pieces in real time, but also predicting future moves by human players. These predictions can be used to create more advanced training systems and offer personalized strategy suggestions, providing valuable information about the evolution of the game. This project seeks to unite two main fronts: the scanning of chessboards and the prediction of moves that imitate human patterns. The proposal combines computer vision techniques, machine learning and realistic three-dimensional (3D) models to make it possible to capture arbitrary configurations of pieces on the board and suggest moves that simulate the style of professional players. In scanning, the approach employed uses images of real boards and 3D models to compose the database. For move prediction, the method applies a model based on convolutional neural networks to classify the moves of the pieces with patterns learned from games of human players. With this, the project seeks to facilitate the understanding of analyses carried out by computers, bringing a more human character to the suggested strategies.
Examination Board
Headlines:
Hélio Pedrini IC / UNICAMP
Alexandre Gonçalves Silva INE / UFSC
André Santanché IC / UNICAMP
Substitutes:
Helena de Almeida Maia IC / UNICAMP
Fátima de Lourdes dos Santos Nunes Marques EACH / USP