DRIFT open dataset: A drone-derived intelligence for traffic analysis in urban environment

TL;DR

The DRIFT paper presents a drone-derived, open traffic dataset collected over nine interconnected urban intersections in Daejeon, Korea, totaling 81,699 vehicle trajectories at ~250 m altitude. It details a fully automated pipeline using orthophoto matching, frame stabilization, and polygon-based OB B detection (YOLOv11) with ByteTracker, achieving high detection accuracy ($mAP@50 \,\approx\,0.994$) and trajectory continuity ($\approx\,96.98\%$). The dataset enables micro-, meso-, and macro-scale traffic analyses and is complemented by open-source models and tools hosted on GitHub, supporting immediate use without heavy preprocessing. DRIFT’s strength lies in continuous network coverage across multiple intersections, enabling analysis of lane changes, TTC, flow-density relationships, and congestion propagation, with potential applications in traffic management, simulation, and digital twins. Limitations include its corridor-focused scope and dependence on weather and visibility, but the openly available resources and preprocessing rigor promote reproducible, data-driven urban mobility research.

Abstract

Reliable traffic data are essential for understanding urban mobility and developing effective traffic management strategies. This study introduces the DRone-derived Intelligence For Traffic analysis (DRIFT) dataset, a large-scale urban traffic dataset collected systematically from synchronized drone videos at approximately 250 meters altitude, covering nine interconnected intersections in Daejeon, South Korea. DRIFT provides high-resolution vehicle trajectories that include directional information, processed through video synchronization and orthomap alignment, resulting in a comprehensive dataset of 81,699 vehicle trajectories. Through our DRIFT dataset, researchers can simultaneously analyze traffic at multiple scales - from individual vehicle maneuvers like lane-changes and safety metrics such as time-to-collision to aggregate network flow dynamics across interconnected urban intersections. The DRIFT dataset is structured to enable immediate use without additional preprocessing, complemented by open-source models for object detection and trajectory extraction, as well as associated analytical tools. DRIFT is expected to significantly contribute to academic research and practical applications, such as traffic flow analysis and simulation studies. The dataset and related resources are publicly accessible at https://github.com/AIxMobility/The-DRIFT.

Figures (15)

• Figure 1: Drone-derived aerial views in nine sites in our experiment (Sties A to I). The green masks indicate the designated RoI for each site.
• Figure 2: Synchronized drone flight periods across nine sites (A–I). Blue shading marks temporally synchronized intervals in our data collections, with a total synchronized duration of approximately 51 minutes and 43 seconds.
• Figure 3: Example of orthophoto in Site H
• Figure 4: Feature detection and matching results between $I_r$ and $I_t$. Feature points were detected using ORB, and corresponding features were matched using the Brute-Force Matcher. Green lines: matched feature points, red lines: failed points.
• Figure 5: Transformation process using $H_{\text{GeoAlign}}$ when feature matching fails. Blue regions: RoIs of $I_r$, red dashed regions: RoI of $I_t$.