Songdo Traffic and Songdo Vision datasets are now publicly accessible at:
- Songdo Traffic Dataset: https://doi.org/10.5281/zenodo.13828384
- Songdo Vision Dataset: https://doi.org/10.5281/zenodo.13828408
AIxMobility Lab, in collaboration with EPFL LUTS Lab, conducted a large-scale multi-drone experiment in the Songdo International District from October 3 to October 7, 2022.
We developed an advanced computer vision algorithm to extract geo-referenced trajectories from drone videos.
The Songdo Traffic dataset includes approximately 700,000 geo-referenced vehicle trajectories, while the Songdo Vision dataset features 300,000 annotated vehicle bounding boxes captured across 20 intersections in the Songdo International District.