Other
RADDet
Radar spectral object detection for dynamic road users
Stationary roadside radar spectra and rectified stereo images with six-class range-azimuth-Doppler and bird-eye-view detection labels.
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
- 3
- Ground truth
- Not reported
- 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.
Not reported
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 maintained release lists sensors_para with stereo rectification matrices/maps and a stereo-to-radar registration matrix; raw calibration sessions are not established.
Known or intentional limitations
Reported constraints and characteristics to check before using the dataset.
The 10,158 samples are frames, not independent capture sessions. Collection used a stationary roadside rig in sunny conditions.
The third RAD axis is Doppler, not geometric elevation. Six-coordinate radar boxes are not spatial 3D cuboids; Cartesian detections are bird-eye-view boxes.
Stereo/radar association and human correction can leave missed objects and class imbalance. Benchmark detection scores are not label-accuracy guarantees.
Original Google Drive links closed. The maintained repository lists replacement access and separate ADC data; current completeness of historical supplemental instance masks was not verified.
Sensor frame rates, exact stereo dimensions, capture-session count and synchronization accuracy remain unknown; the radar carrier-frequency configuration is not a frame cadence.
Cameras
Camera 1
- Model
- The Imaging Source DFK 33UX273
- Setup
- Stereo
- Count
- 1
- Effective cameras
- 2
- Modality
- Rgb
- Resolution
- Not reported
- Rate
- Not reported
- Shutter
- Not Reported
- Lens type
- Not Reported
- HFOV
- Not reported
- VFOV
- Not reported
- Notes
- Two rectified stereo images per frame are explicitly listed in the maintained release; exact released dimensions and cadence are not established.
IMUs
Not reported
GNSS
Not reported
LiDAR
Not reported
Radar
Texas Instruments AWR1843-BOOST
- Count
- 1
- Released data
- Spectra / dense tensors
- Frame rate
- Not reported
- Doppler velocity
- Available
- Elevation angle
- Not available
- Configured range
- 50 m
- Field of view
- Not reported
- File format
- NumPy RAD tensors
- Notes
- Complex RAD tensor 256×256×64 after FFT/zero-padding; not a spatial xyz/elevation radar cube.
Additional sensors
Not reported
Annotations
Reported annotation availability and task support.
- Availability
- Available
- Format
- Range/azimuth/Doppler-domain boxes and Cartesian bird-eye-view boxes; automatic association plus human correction
- Class count
- 6
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.
@misc{zhang2021raddet,
author = {Zhang, Ao and Nowruzi, Farzan Erlik and Laganiere, Robert},
title = {RADDet: Range-Azimuth-Doppler based Radar Object Detection for Dynamic Road Users},
year = {2021},
eprint = {2105.00363},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2105.00363v1}
}Dataset corrections
Notice a specific issue after reviewing the page?
Report an inaccuracy