2019/04/30 by Braden Hurl, Krzysztof Czarnecki, Hurl, Braden +3
Computer Science · Engineering · #Advanced Neural Network Applications #Autonomous Vehicle Technology and Safety #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Robotics (cs.RO) #Robotics and Sensor-Based Localization
paper · pdf · doi:10.48550/arxiv.1905.00160
openalex publication_date 2019/04/30 · openalex created_date 2022/07/29 · openalex updated_date 2026/07/28
We introduce the Precise Synthetic Image and LiDAR (PreSIL) dataset for\nautonomous vehicle perception. Grand Theft Auto V (GTA V), a commercial video\ngame, has a large detailed world with realistic graphics, which provides a\ndiverse data collection environment. Existing works creating synthetic LiDAR\ndata for autonomous driving with GTA V have not released their datasets, rely\non an in-game raycasting function which represents people as cylinders, and can\nfail to capture vehicles past 30 metres. Our work creates a precise LiDAR\nsimulator within GTA V which collides with detailed models for all entities no\nmatter the type or position. The PreSIL dataset consists of over 50,000 frames\nand includes high-definition images with full resolution depth information,\nsemantic segmentation (images), point-wise segmentation (point clouds), and\ndetailed annotations for all vehicles and people. Collecting additional data\nwith our framework is entirely automatic and requires no human annotation of\nany kind. We demonstrate the effectiveness of our dataset by showing an\nimprovement of up to 5% average precision on the KITTI 3D Object Detection\nbenchmark challenge when state-of-the-art 3D object detection networks are\npre-trained with our data. The data and code are available at\nhttps://tinyurl.com/y3tb9sxy\n