prakharbh/CrossViewUrbanTrafficDataset

Cross-View Urban Traffic Dataset

Dataset Summary

The Cross-View Urban Traffic Dataset (CVUTD) is a benchmark for cross-view urban traffic perception built from synchronized ego-centric bicycle videos and aerial drone videos recorded at real urban intersections in Regensburg, Germany.


The dataset is designed to support two linked tasks:

1. 1.Cross-view identity matching between street-view and drone-view object tracks

1. 1.Ego-to-BEV prediction using aerial supervision

The benchmark focuses on intersection-centric traffic analysis, where local interactions, identity preservation, and global spatial structure must be reasoned about jointly across views.

Supported Tasks and Leaderboards

Task 1: Cross-view identity matching

Given synchronized street-view and drone-view object tracks, predict which street-view track corresponds to which drone-view track.

Typical metrics:

• —Track Precision / Recall / F1

• —ID Precision / Recall / IDF1

• —Frame assignment accuracy

• —Near/Far breakdown

• —Stability

• —ID switches

• —Consistency

Task 2: Ego-to-BEV prediction

Given an ego-centric street-view image or sequence, predict the spatial arrangement of traffic participants in a shared bird’s-eye-view frame.

Typical metrics:

• —ADE

• —FDE

• —ALE

• —ALgE

• —PCK@1m / PCK@2m

• —mIoU / IoU@thresholds

Languages

This dataset is primarily visual and geometric. Language metadata is included only for the English documentation and labels.

Dataset Structure

A typical scene contains:

• —synchronized street-view video

• —synchronized drone-view video

• —street-view detection/tracking CSV

• —drone-view detection/tracking CSV

• —verified cross-view correspondences

• —processed wedge-filtered matching artifacts

• —alignment metadata for BEV evaluation

Example structure:

CrossViewUrbanTrafficDataset/

  README.md

  LICENSE

  scene_manifest.csv (found on github)

  scenes/

    scene_01_01/

      street_video.mp4

      drone_video.mp4

      street_detections.csv

      drone_detections.csv

      gt_pairs.csv

      gt_audit.csv

      outputs/

        wedge_export/

          street_wedge_manifest.csv

          drone_wedge_manifest.csv

          frame_matches.csv

          track_mapping.csv

Citation

If you use this dataset in your research, please cite:

@misc{bhardwaj2026crossviewurbantrafficdataset,
      title={Cross-View Urban Traffic Dataset: Drone-Supervised Ground Truth for Monocular Bird's-Eye View Localization}, 
      author={Prakhar Bhardwaj and Simone Weikl and Kilian Mang and Elia Jonas Sandtner},
      year={2026},
      eprint={2606.07708},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2606.07708}, 
}