The JKU DORA Traffic Dataset

Abstract

Development and testing of Advanced Driver Assistance Systems (ADAS) is largely based on models and simulations, but real data are indispensable for many reasons – to determine the relevant scenarios, to establish a connection between the results of the simulations and the real situation and of course as elements to set up realistic models. Using data, however, is not trivial, as not all data are informative, and even extensive data sets are often incomplete. Indeed, data is not automatically information, and the richness of the data sets is more important than their size. Data should be diverse not only with respect to different scenarios but also geographically in order to be not biased towards a specific location. In this paper, we present a new aerial-view dataset “Drone Over Roads” (DORA) of highway exits and entrances. The data have been collected by the Johannes Kepler University Linz (JKU), Austria and Italy and contain positions, velocities, and accelerations (in both local and global coordinate systems) of cars, vans, and trucks for Austria and additionally for motorbikes and buses in Italy. The uniqueness of the dataset consists not only in the measurements in different countries, but also in the flight height in Austria, where the recordings have been taken from 300 meters altitude allowing to observe an over 600 meters long section of the road. For non-commercial use, the dataset is available free of charge at the IEEE DataPort.

Keywords

• Dataset
• drone data
• highway entrances/exits
• traffic