Udacity Dataset is a set of annotated image datasets, mainly used for autonomous driving. It contains two sub-datasets with 9423 and 15000 images respectively, with a resolution of 1920 * 1200. This dataset was prepared by Udacity for the autonomous driving algorithm competition in 2016. It carefully annotates continuous video images and mainly divides them into categories such as cars and pedestrians. The related paper is "Learning a Driving Simulator".