Using Deep Convolutional Neural Networks to Detect Rendered Glitches in Video Games"
Machine Learning
Graphical errors are often hard to spot by eye during game testing. This paper presents a method for using Deep Convolutional Neural Networks (DCNNs) to detect common visual glitches in video games. The main use of this work is the partial automatization of graphical testing in the final stages of video game development.
Developing video games involves many steps, starting from the concept, to the final release. Often there are hundreds of developers and artists involved when creating a modern game. In this complex process, plenty of bugs can be introduced, and many of them having a negative effect on the rendered images. We refer to these graphical bugs as glitches.
Graphical glitches can occur at several stages: when updating the asset database (resulting in missing textures), updating the codebase (resulting in textures being corrupted), updating drivers, cross-platform development, and so on. Since graphics are one of the primary components of any video game, it is of high importance to assure the absence of glitches or malfunctions that otherwise may negatively affect the player’s experience.
This paper will also be presented at the 16th AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment, October 19-23, 2020.
Authors: Carlos García Ling, Konrad Tollmar, Linus Gisslén
Download the Paper "Using Deep Convolutional Neural Networks to Detect Rendered Glitches in Video Games"
Download the paper as PDF (5.2 MB).
