Dataset for multi-perspective traffic video analysis

Multi-angle video recordings are of significant interest as they provide comprehensive and synchronized data from multiple perspectives. This enables advanced analysis and modeling in environments with complex conditions due to their dynamics or the presence of obstacles, which lead to occlusions or loss of information in single-camera setups. This dataset comprises video footage of vehicular and pedestrian activity from three complementary viewpoints: A vehicle-mounted camera, a road side unit-mounted surveillance camera, and a drone-mounted camera. The dataset can be used in several domains including object, event, and activity recognition and tracking, urban analysis and planning, digital twinning, as well as semantic communication. To demonstrate its usefulness, standardized metrics are employed to provide quantitative evidence of the dataset’s quality and reliability in comparison with existing datasets. The multi-modal nature of the footage paves the way for novel research in sensor fusion architectures and multi-scale modeling, critical for next-generation smart urban infrastructures.