Overview
DRIFT is an open dataset and toolchain for drone-based traffic behavior research, connecting vehicle detection, trajectory tracking, and multi-scale traffic analytics across complex urban intersections.
Research Focus
- Drone-video stabilization and orthophoto-aligned trajectory extraction for urban corridors.
- Vehicle detection and tracking using polygon-based oriented bounding boxes with YOLOv11m and ByteTrack.
- Traffic-analysis tools for lane changes, time-to-collision, congestion, flow-density, time-space, and speed-heatmap views.
Technical Keywords
- DRIFT
- Drone Traffic Dataset
- Vehicle Trajectories
- Urban Intersections
- YOLOv11m
- ByteTrack
- Traffic Flow Analysis
Applications
- Benchmarking detection, tracking, and trajectory-prediction methods in dense urban traffic.
- Studying lane-change behavior, near-conflict patterns, and congestion propagation from drone-derived trajectories.
- Preparing a dashboard layer for interactive exploration of sites, trajectories, and traffic-analysis outputs.
Representative Outputs