Multi-Label Perceptual Bug Detection in Video Games using Deep Learning on Gameplay Footage
This research addresses the challenge of automatically detecting multiple visual bugs in video game footage, which is difficult because bugs can appear simultaneously and require understanding visual details. The study proposes a ResNet-BiLSTM deep learning model to perform multi-label perceptual bug detection. The model achieved an F1 score of 85.78% on a new dataset containing combinations of five bug types, suggesting temporal modeling is key for accurate video bug identification.