Automotive
OPV2V
Vehicle-to-vehicle cooperative perception
Simulated multi-agent camera/LiDAR scenes pair with vehicle boxes, BEV maps and replayable driving logs.
Access links
Primary places to inspect, download, read about, or reproduce the dataset.
Quick facts
- Vehicle type
- Automotive
- Environment
- Outdoor
- Data origin
- Simulation
- Sensor count
- 5
- 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
|
|---|---|---|---|---|---|---|
| 6DoF | Other | CARLA simulator ego pose | Not reported | 10 Hz | Full |
Reference provenance
Derived
|
|
||||||
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
Per-timestamp YAML metadata includes camera intrinsics and camera-to-LiDAR extrinsics; simulator configuration is supplied for replay. These are synthetic parameters, not physical calibration sessions.
Known or intentional limitations
Reported constraints and characteristics to check before using the dataset.
The 73 scenarios contain 11,464 collective time steps and two to seven participating vehicles. Four cameras and one LiDAR describe each agent, not a fixed aggregate fleet rig.
Data is synthetic, including the reconstructed Culver City town. Physical sensors used to build that digital town are not additional released input streams.
Per-agent simulator poses and calibration matrices are metadata. Separately released raw GNSS/IMU observations are not established; configured simulator devices do not increase the sensor inventory.
BEV semantics are not image-plane masks. Visibility-qualified vehicle boxes and simulated noise parameters are not independently measured reference accuracy.
Communication radius, delay and compression settings belong to experiments, not measured radio logs. Replay can add sensors beyond the original release.
Cameras
Camera 1
- Model
- Not reported
- Setup
- Mono
- Count
- 4
- Effective cameras
- 4
- Modality
- Rgb
- Resolution
- 800 x 600
- Rate
- 10 Hz
- Shutter
- Not Reported
- Lens type
- Not Reported
- HFOV
- Not reported
- VFOV
- Not reported
- Notes
- Per-agent inventory; simulated cameras stream 20 Hz but recording is 10 Hz.
IMUs
Not reported
GNSS
Not reported
LiDAR
LiDAR 1
- Model or name
- LiDAR 1
- Count
- 1
- Dimensions
- 3
- Rate
- 10 Hz
- Channels
- 64
- Range
- 120 m
Radar
Not reported
Additional sensors
Not reported
Annotations
Reported annotation availability and task support.
- Availability
- Available
- Format
- 3D vehicle cuboids and per-agent bird-eye-view semantic maps
- Class count
- Not reported
Semantic segmentation
Available
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{xu2021opv2v,
author = {Xu, Runsheng and Xiang, Hao and Xia, Xin and Han, Xu and Li, Jinlong and Ma, Jiaqi},
title = {OPV2V: An Open Benchmark Dataset and Fusion Pipeline for Perception with Vehicle-to-Vehicle Communication},
year = {2021},
eprint = {2109.07644},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2109.07644v5}
}Dataset corrections
Notice a specific issue after reviewing the page?
Report an inaccuracy