FURI | Fall 2023

Video Game-Based Domain Generalization for Computer Vision

This research project seeks to introduce a fresh dimension to the field of Domain Generalization (DG) in computer vision. It aims to achieve this by developing a unique benchmark dataset composed of scenes from video games. Unlike current benchmarks, our dataset encompasses transitional phases in video game graphics. The project encompasses data gathering, data refinement, and manual annotation. After the dataset is assembled, we will establish DG performance benchmarks on our dataset using existing DG techniques. Both datasets and results will be made available to the public, encouraging further exploration and research into Domain Generalization within the context of video game graphics.