ResearchDoom and CocoDoom: Learning Computer Vision with Games
This provides a new tool and dataset for computer vision researchers to simulate and test methods in a controlled gaming environment, but it is incremental as it adapts existing formats and games.
The authors introduced ResearchDoom, a modified Doom game for extracting detailed metadata, and CocoDoom, a dataset of annotated gaming data in MS Coco format, enabling training and evaluation of various computer vision tasks like object detection and tracking.
In this short note we introduce ResearchDoom, an implementation of the Doom first-person shooter that can extract detailed metadata from the game. We also introduce the CocoDoom dataset, a collection of pre-recorded data extracted from Doom gaming sessions along with annotations in the MS Coco format. ResearchDoom and CocoDoom can be used to train and evaluate a variety of computer vision methods such as object recognition, detection and segmentation at the level of instances and categories, tracking, ego-motion estimation, monocular depth estimation and scene segmentation. The code and data are available at http://www.robots.ox.ac.uk/~vgg/research/researchdoom.