Reproduce “Using Deep Convolutional Neural Networks to Detect Rendered Glitches in Video Games” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Carlos Garcia Ling, Konrad Tollmar, Linus Gisslen
A deep convolutional neural network approach was developed and evaluated for automatically identifying graphical glitches in video games. The method achieved an average validation accuracy of 86.8% in classifying normal frames or one of four specific texture-related glitch types, demonstrating a viable path to automate a component of game quality assurance.
Key insights
- Pre-training DCNNs on ImageNet significantly enhances performance for glitch detection, indicating transferability of features from natural images to game visuals.
- The design of placeholder textures influences detection accuracy; patterned placeholders are more easily classified than plain ones, suggesting practical guidance for asset pipeline design.
- ShuffleNetV2 emerges as an optimal architecture, balancing high accuracy (86.8%) with real-time computational efficiency (60 frames per second) for this specific task.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.