Other
R-LiViT
Roadside vulnerable road user detection and tracking
Roadside LiDAR, RGB and thermal observations at three German intersections in day and night, with 10,000 labeled point-cloud frames and 2,400 aligned RGB-T pairs.
Access links
Primary places to inspect, download, read about, or reproduce the dataset.
Quick facts
- Vehicle type
- Other
- Environment
- Outdoor
- Data origin
- Real-world
- Sensor count
- 4
- Ground truth
- Available
- Calibration
- Not reported
- 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
|
|---|---|---|---|---|---|---|
| 3D position | Manual annotation | Not reported | Not reported | 5 Hz | Partial · LiDAR Objects Within 50 M |
Reference provenance
Derived
|
|
||||||
Calibration and synchronization
Reported calibration level, reproducibility signals, and supporting notes.
- Level
- Not reported
- Processed parameters
- Not reported
- Raw calibration data
- Not reported
- Calibration targets
- RGB/LiDAR: aluminum checkerboard, 6x8 tiles of 14x14 cm. Thermal/RGB: 5x7 circle target, 95 mm diagonal spacing and 55 mm circles.
Calibration data
Calibration reproducibility details are not reported.
Notes
Thermal-to-RGB homography estimated from the target; thermal frames projected and padded. Reported mean reprojection error 0.038 has no explicit unit. Procedure does not establish delivery of parameters or raw target observations.
Known or intentional limitations
Reported constraints and characteristics to check before using the dataset.
Occasional temporal offsets between RGB-T and LiDAR.
Independent modality labels and differing fields of view produce inconsistent object coverage; no cross-modal track IDs.
LiDAR labels stop at 50 m and omit e-scooters; RGB-T cyclists have separate person/bicycle boxes.
Clear-weather spring collection with operator-selected active traffic scenarios.
59.3% of 2D boxes are small objects; valid 3D boxes contain approximately 110 points on average.
Cameras
rgb
- Model
- Hikvision DS-2TD2628-3/QA
- Setup
- Mono
- Count
- 1
- Effective cameras
- 1
- Modality
- Rgb
- Resolution
- Not reported
- Rate
- Not reported
- Shutter
- Not Reported
- Lens type
- Not Reported
- HFOV
- Not reported
- VFOV
- Not reported
- Notes
- Native RGB capture is 1280x720 at 5 Hz; RGB-T annotation subset is 1.25 FPS. Identifying regions are blurred. Benchmark 800x600 crops are experimental preprocessing. Cross-modal temporal offsets can remain.
thermal
- Model
- FLIR ADK
- Setup
- Mono
- Count
- 1
- Effective cameras
- 1
- Modality
- Thermal
- Resolution
- Not reported
- Rate
- Not reported
- Shutter
- Not Reported
- Lens type
- Not Reported
- HFOV
- Not reported
- VFOV
- Not reported
- Notes
- Native thermal capture is 640x512 at 60 Hz. Images are projected into RGB coordinates and padded; the aligned RGB-T subset has 2,400 image pairs at 1.25 FPS. Native resolution is not asserted as delivered aligned geometry.
IMUs
Not reported
GNSS
Not reported
LiDAR
Ouster OS1 BH
- Model or name
- Ouster OS1 BH
- Count
- 1
- Dimensions
- 3
- Rate
- Not reported
- Channels
- 64
- Range
- Not reported
RoboSense Bpearl
- Model or name
- RoboSense Bpearl
- Count
- 1
- Dimensions
- 3
- Rate
- Not reported
- Channels
- 32
- Range
- Not reported
Radar
Not reported
Pitot tube
Not reported
Additional sensors
Not reported
Annotations
Reported annotation availability and task support.
- Availability
- Available
- Format
- Not reported
- Class count
- Not reported
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.
Not reported
Dataset corrections
Notice a specific issue after reviewing the page?
Report an inaccuracy