Code and dataset of paper IEEE IV 2025 "Framework and Multi-modal Dataset for Roadwork Zone Detection and Geo-localization"
The RZDG (Roadwork Zone Detection and Geo-localization) dataset is a multi-modal dataset designed to support research in roadwork zone perception, detection, tracking and geo-localization. It contains both real-world (RZDG-real) and simulated (RZDG-sim) data, structured in the KITTI format, and supports tasks such as:
- Image semantic segmentation
- 3D object detection and tracking
- Geo-localization of roadwork zones
The download is available at: RZDG Download Link
- Objects: 419 road beacons, 254 barriers
- Scenarios: 203 distinct roadwork zone scenarios across 8 simulated towns
- Frames: 8591
- Objects: 3643 road beacons, 8837 barriers
- Scenarios: 28 distinct roadwork zone scenarios
- Frames: 1357
The overall structure of the dataset is as follows:
RWDG_dataset
├── ImageSets
├── kitti_gt_database
├── testing
│ ├── calib
│ ├── image_0
│ └── lidar
└── training
├── calib
├── image_0
├── label_0
└── lidar
We adopt MMDetection3D for 3D object detection and AB3DMOT for 3D object tracking.
To adapt these tools to the RZDG dataset, we provide custom modifications in the form of patch files located in the patches/ directory.
| File | Target Repository | Description |
|---|---|---|
patches/mmdetection3d.patch |
MMDetection3D | Modifications to support RZDG dataset format, object classes, and evaluation setup |
patches/AB3DMOT.patch |
AB3DMOT | Tracker changes to handle geo-localization and RZDG-specific object cues |
Use following commands to apply patches to the Original Repositories
cd path/to/mmdetection3d
git apply ../patches/mmdetection3d.patch
cd path/to/AB3DMOT
git apply ../patches/AB3DMOT.patchThis project is built upon the excellent open-source frameworks MMDetection3D and AB3DMOT. We sincerely thank the developers and maintainers of these repositories for their contributions to the research community.
@inproceedings{yan2025framework,
title={Framework and Multi-modal Dataset for Roadwork Zone Detection and Geo-localization},
author={Yan, Zhiran and Xin, Yutong and Shenoi, S Shyam and Song, Rui and Elger, Gordon},
booktitle={2025 IEEE Intelligent Vehicles Symposium (IV)},
pages={1758--1765},
year={2025},
organization={IEEE}
}