TIAS: Ein Verkehrsinfrastrukturdatensatz aus der Luft zur Untersuchung des Verkehrs in städtischen Gebieten

Kurzfassung

For the 2nd Forum GEO.KI, we want to present the TIAS dataset, discuss the challenges behind its creation, and show possible applications. TIAS is a novel dataset consisting of aerial images and high quality labels of traffic areas. By providing detailed, fine-grained annotations, it accurately reflects urban scenarios from a transportation perspective. The traffic areas are classified into nine classes: road, access way, bikeway, footway, keep-out area, parking area, railroad bed, road shoulder, and water. To preserve the topological nature of the transportation network, an attribute indicates whether areas are shared by two or more traffic participants. The dataset enables applications such as the reconstruction of the traffic network for motorized vehicles, bicycles, pedestrians, and rail-based traffic, or the identification of hazardous areas (e.g. for automated vehicles and traffic safety analysis).