List of Datasets

Automotive

KITTI-360

Joint 2D/3D semantics, novel view synthesis and semantic SLAM

A 73.7 km driving collection pairs stereo/fisheye images and two LiDAR types with dependent poses and coherent semantic/instance products.

Metadata: Complete Metadata source documented Released 2020

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

Quick facts

Vehicle type
Automotive
Environment
Outdoor
Data origin
Real-world
Sensor count
8
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

2 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.
6DoF Structure from motion SfM with GPS and LiDAR regularization Not reported Not reported Partial · Released Pose Frames 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
Reference notes
SfM with GPS regularization and LiDAR loop closures reuses recorded inputs. poses.txt omits frames when movement since the last valid pose is below a threshold; no per-frame independent truth is promised.

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

Official calibration directory provides perspective/fisheye intrinsics and camera-to-pose, camera-to-Velodyne and SICK-to-Velodyne extrinsics. Raw calibration recordings are not established.

Known or intentional limitations

Reported constraints and characteristics to check before using the dataset.

Eleven continuous driving sequences cover 73.7 km. Overlapping annotation windows and image counts are different units; raw-scan/image totals differ across the paper and maintained release descriptions.

Pose estimates reuse image, GPS and LiDAR measurements and omit some frames below a movement threshold. Numerical independent pose accuracy is not established.

Native 37-class labels, 19-class training/evaluation conventions and task-specific exclusions differ. Dense 2D products concern perspective views; confidence maps qualify transferred labels.

Test labels and selected accumulated clouds are withheld. The right-perspective label release is a later documented update; the paper appendix contains an unresolved drop-rate protocol inconsistency.

Downloads require registration and a stated research purpose. Exact sensor models/rates, raw calibration sessions and independent timing/reference errors remain unknown.

Cameras

Camera 1

Model
Not reported
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
Front perspective stereo, 90-degree view and 60 cm baseline.

Camera 2

Model
Not reported
Setup
Mono
Count
2
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 side 180-degree fisheye views; dense 2D labels apply to perspective imagery.

IMUs

OXTS navigation unit IMU

Model
Not reported
Count
1
Accel / gyro
Not reported

GNSS

OXTS navigation unit

Model or name
OXTS navigation unit
Count
1
Rate
Not reported
GNSS type
GPS_IMU
Position output
Available
Velocity output
Available
Heading output
Available

LiDAR

Velodyne HDL-64E

Model or name
Velodyne HDL-64E
Count
1
Dimensions
3
Rate
Not reported
Channels
Not reported
Range
Not reported

SICK LMS 200

Model or name
SICK LMS 200
Count
1
Dimensions
2
Rate
Not reported
Channels
Not reported
Range
Not reported

Radar

Not reported

Additional sensors

Not reported

Automotive

KITTI

Outdoor automotive dataset with stereo cameras, LiDAR, IMU/GNSS, and GPS/IMU ground truth.

Annotations

Reported annotation availability and task support.

Availability
Available
Format
3D bounding primitives, 2D semantic/instance maps and 3D point labels with confidence; 37 native versus 19 benchmark training classes
Class count
37
Semantic segmentation Available
Instance segmentation Available
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.

Citation key
liao2022kitti360

BibTeX

@article{liao2022kitti360,
  author = {{Yiyi Liao} and {Jun Xie} and {Andreas Geiger}},
  title = {KITTI-360: A Novel Dataset and Benchmarks for Urban Scene Understanding in 2D and 3D},
  year = {2022},
  journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},
  eprint = {2109.13410},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2109.13410v2}
}

Dataset corrections

Notice a specific issue after reviewing the page?

Report an inaccuracy