Abstract
Gamers use polyhedral dice that come with 4 sides (D4), 6 sides (D6), 8 sides (D8), 10 sides (D10), 12 sides (D12) and 20 sides (D20). All dice are unfair, some more than others. The goal of this project was to develop a machine learning and computer vision solution for the interpretation of dice rolls. When combined with an automated dice roller it would facilitate the study of dice unfairness. In the machine learning literature, Convolutional Neural Networks (CNNs) are the preferred method of computer classification of images. The CNN convolves images with different weights and biases through multiple layers to produce an end array with odds of each number on the die. Using CNNs we were able to obtain interpretation accuracies ranging from 95.10% to 99.76%. Thousands of images were used for training, and thousands of separate images used for validation.
Disciplines
Engineering | Engineering Science and Materials | Materials Science and Engineering
Recommended Citation
Wimsatt, Hunter; Panzade, Aarohi; Shankar, Kaaustaaub; and Campbell, Warren, "Using Machine Learning to Interpret Dice Rolls" (2021). SEAS Faculty Publications. Paper 5.
https://digitalcommons.wku.edu/seas_faculty_pubs/5