List of Datasets

Automotive

DSEC

Stereo event vision for driving

Driving recordings with stereo event/RGB cameras, lidar, a released IMU stream and partial depth/disparity references. Later flow and semantic labels are available.

Metadata: Complete Metadata curator reviewed Released 2021

Primary places to inspect, download, read about, or reproduce the dataset.

Supplementary

Not reported

Quick facts

Vehicle type
Automotive
Environment
Outdoor
Data origin
Real-world
Sensor count
7
Ground truth
Available
Calibration
Parameters Only
Annotations
Available

Sensor overview

High-level sensor availability before the detailed sensor records below.

Cameras

4 reported

IMUs

1 reported

GNSS

1 reported

LiDAR

1 reported

Radar

Not reported

Magnetometers

Not reported

Barometers

Not reported

UWB

Not reported

Ground truth

Reference scope, method, rate, coverage, and provenance.

Scope Method Reference system Count Rate Coverage Reference provenance
External
Measured using equipment independent of the platform’s normal onboard sensor pipeline.
Derived
Computed primarily from the dataset’s own recorded sensors.
Hybrid
Combines independent reference equipment with onboard measurements.
Other Other Not reported Not reported 10 Hz Partial · Valid Left RGB/Event Reference Pixels at Up to 50 M Reference provenance
External
Measured using equipment independent of the platform’s normal onboard sensor pipeline.
Derived
Computed primarily from the dataset’s own recorded sensors.
Hybrid
Combines independent reference equipment with onboard measurements.
Derived
Accuracy notes
Nominal LiDAR precision does not establish the final projected/reference error.
Reference notes
Accumulated LiDAR using LIO-Mapping poses, with motion/occlusion and RGB SGM filtering. Depth/disparity and two left-camera projections are one coherent reference source.
3D position RTK GNSS u-blox RTK · ZED-F9P Not reported 10 Hz Unknown Reference provenance
External
Measured using equipment independent of the platform’s normal onboard sensor pipeline.
Derived
Computed primarily from the dataset’s own recorded sensors.
Hybrid
Combines independent reference equipment with onboard measurements.
External
Accuracy notes
Nominal 1 cm receiver precision is not a collection-wide realized error.
Reference notes
No independent pose/attitude reference is established.

Calibration and synchronization

Reported calibration level, reproducibility signals, and supporting notes.

Level
Parameters Only
Processed parameters
Available
Raw calibration data
Not reported
Calibration targets
checkerboard

Parameters only

Processed calibration parameters are reported, but raw calibration data is not reported.

Notes

Release includes camera and lidar/IMU calibration files. Complete raw calibration recordings are not established.

Known or intentional limitations

Reported constraints and characteristics to check before using the dataset.

Reference coverage is sparse and depends on RGB SGM acceptance, illumination and dynamic-object filtering.

Night event noise, contrast-threshold changes and strong-flash saturation affect event statistics.

RGB and event stereo pairs have different baselines and optical centers.

Training references are public; most test references are withheld. Later flow labels cover only selected timestamps and subsets of sequences.

Semantic labels are generated from warped frame-camera images; the last 40 rows are cropped. Separate 11- and 19-class label sets are available.

Released IMU hardware and rates and independent RTK accuracy remain unknown.

Cameras

Camera 1

Model
FLIR Blackfly S BFS-U3-16S2C-CS
Setup
Stereo
Count
1
Effective cameras
2
Modality
Rgb
Resolution
1440 x 1080
Rate
20 Hz
Shutter
Global
Lens type
Not Reported
HFOV
Not reported
VFOV
Not reported
Notes
51 cm RGB baseline; up to 15 ms night exposure. One system represents two imagers.

Camera 2

Model
Prophesee Gen3.1
Setup
Stereo
Count
1
Effective cameras
2
Modality
Event
Resolution
640 x 480
Rate
Not reported
Shutter
Not Reported
Lens type
Not Reported
HFOV
Not reported
VFOV
Not reported
Notes
Monochrome asynchronous events, PPS3MVCD sensors, 60 cm baseline. The 20 Hz timing markers are not an event sampling rate.

IMUs

Released IMU stream

Model
Not reported
Count
1
Accel / gyro
Not reported

GNSS

u-blox ZED-F9P

Model or name
u-blox ZED-F9P
Count
1
Rate
10 Hz
GNSS type
RTK
Position output
Available
Velocity output
Not reported
Heading output
Not reported

LiDAR

Velodyne VLP16

Model or name
Velodyne VLP16
Count
1
Dimensions
3
Rate
10 Hz
Channels
16
Range
100 m

Radar

Not reported

Additional sensors

Not reported

Annotations

Reported annotation availability and task support.

Availability
Available
Format
Events in HDF5; rectified images and disparity/flow masks in PNG. Later left-event-view semantic labels use separate 11- and 19-class taxonomies; no single merged class list.
Class count
Not reported
Semantic segmentation Available
Instance segmentation Not reported
Object detection Not reported
Optical flow Available
Depth ground truth Available

Citation

Citation key and BibTeX kept at the end of the page for reference.

Citation key
gehrig2021dsec

BibTeX

@misc{gehrig2021dsec,
  title = {DSEC: A Stereo Event Camera Dataset for Driving Scenarios},
  author = {Mathias Gehrig and Willem Aarents and Daniel Gehrig and Davide Scaramuzza},
  year = {2021},
  eprint = {2103.06011v1},
  archivePrefix = {arXiv}
}

Dataset corrections

Notice a specific issue after reviewing the page?

Report an inaccuracy