Project Abstract

Computer vision is at the forefront of research in the machine learning world, classification pays a crucial role in this to define what it actually is a computer can see. This project presents a full pipeline from a camera observing a chess board to suggesting the next best move.This is achieved by processing the information provided by a camera and converting it into a 3-Dimensional representation that can be understood by a computer. This point cloud is then processed to identify all the pieces and their positions on the board. Finally these pieces are then identified using machine learning models, this information is then processed through a chess engine to produce the next best possible move for that position.This paper presents multiple different machine learning architectures for classification comparison to find the ones most proficient. Our final model managed to achieve a 99.53% accuracy on a validation set of pieces.To summarise this paper presents a comparison of point cloud classification models and releases the dataset that they were trained on for future chess related computer vision research.

Keywords: Computer Vision, Deep Neural Networks, Point cloud

 Conference Details

Session: Poster Session A at Poster Stand 94

Location: Sir Stanley Clarke Auditorium at Tuesday 7th 13:30 – 17:00

Markers: Gary Tam, C�cilia Pradic

Course: BSc Computer Science, 3rd Year

Future Plans: I’m looking for work