Automotive
CRUW ROD2021 Camera-Radar Dataset
Radar object detection using the downloadable CRUW ROD2021 subset
Fifty synchronized 30 Hz camera-radar sequences across parking-lot, campus-road, city-street, and highway scenarios, with range-azimuth point labels for radar object detection.
Access links
Primary places to inspect, download, read about, or reproduce the dataset.
Quick facts
- Vehicle type
- Automotive
- Environment
- Outdoor
- Data origin
- Real-world
- Sensor count
- 2
- Ground truth
- Available
- Calibration
- Parameters Only
- Annotations
- Available
Sensor overview
High-level sensor availability before the detailed sensor records below.
Ground truth
Reference scope, method, rate, coverage, and provenance.
| Scope | Method | Reference system | Count | Rate | Coverage |
Reference provenance
|
|---|---|---|---|---|---|---|
| 2D | Other | Camera-radar fusion annotator with human labeling | Not reported | Not reported | Partial · Training Set |
Reference provenance
Hybrid
Accuracy and notes
|
Calibration and synchronization
Reported calibration level, reproducibility signals, and supporting notes.
- Level
- Parameters Only
- Processed parameters
- Available
- Raw calibration data
- Not reported
- Calibration targets
- Not reported
Parameters only
Processed calibration parameters are reported, but raw calibration data is not reported.
Notes
The CRUW devkit supplies sensor configurations, calibration parameters, and camera-radar coordinate mappings; RGB images are undistorted or rectified before release.
Known or intentional limitations
Reported constraints and characteristics to check before using the dataset.
The downloadable release is the 50-sequence ROD2021 subset rather than the complete 464-sequence CRUW corpus described by the dataset paper.
Testing sequences release radar RF images only; camera images and human ground-truth labels are withheld for evaluation.
Only four uniformly selected chirps per frame are released from the 255 captured chirps.
Point annotations do not provide object dimensions or other attributes.
Cameras
Camera 1
- Model
- Not reported
- Setup
- Mono
- Count
- 1
- Effective cameras
- 1
- Modality
- Rgb
- Resolution
- Not reported
- Rate
- 30 Hz
- Shutter
- Not Reported
- Lens type
- Not Reported
- HFOV
- Not reported
- VFOV
- Not reported
- Notes
- RGB images are released for the 40 training sequences; the 10 downloadable testing sequences provide radar input only.
IMUs
Not reported
GNSS
Not reported
LiDAR
Not reported
Additional sensors
Radar
- Name
- 77 GHz FMCW millimeter-wave radar
- Count
- 1
- Frequency ghz
- 77
- Rate
- 30 Hz
- Notes
- Released data are normalized range-azimuth RF images spanning 0-25 m and +/-60 degrees. Four uniformly selected chirps are released from 255 captured chirps per frame.
Annotations
Reported annotation availability and task support.
- Availability
- Available
- Format
- Point annotations with class, range in metres, and azimuth in radians
- Class count
- [car pedestrian cyclist]
Semantic segmentation
Not reported
Instance segmentation
Not reported
Object detection
Available
Optical flow
Not reported
Depth ground truth
Not reported
Citation
Citation key and BibTeX kept at the end of the page for reference.
@inproceedings{Wang_2021_WACV, author={Wang, Yizhou and Jiang, Zhongyu and Gao, Xiangyu and Hwang, Jenq-Neng and Xing, Guanbin and Liu, Hui}, title={RODNet: Radar Object Detection Using Cross-Modal Supervision}, booktitle={Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)}, month={January}, year={2021}, pages={504--513}}Dataset corrections
Notice a specific issue after reviewing the page?
Report an inaccuracy