List of Datasets

UGV

NAVER Labs

Crowded indoor spaces

Indoor ground-robot dataset for crowded spaces with multiple RGB camera systems and LiDAR/SfM ground truth.

Metadata: Complete Metadata source documented Released 2021

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

Quick facts

Vehicle type
UGV
Environment
Indoor
Data origin
Real-world
Sensor count
14
Ground truth
Available
Calibration
Parameters Only
Annotations
Available

Sensor overview

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

Cameras

10 reported

IMUs

Not reported

GNSS

Not reported

LiDAR

2 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 LiDAR SLAM Not reported Not reported Not reported Partial · Public Splits 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 and notes
Reference notes
LiDAR pose-graph SLAM initializes each platform trajectory. Test-set poses are withheld for benchmark evaluation.
6DoF Structure from motion Not reported Not reported Not reported Partial · Public Splits 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 and notes
Accuracy notes
Final mean reprojection errors are below 1.5 pixels for all five environments.
Reference notes
Structure-from-motion optimization jointly refines asynchronous image poses, 3D points, and calibration parameters. Test-set poses are withheld.

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 LiDAR-to-LiDAR transform is calibrated with ICP. Camera intrinsics and rotational camera-to-platform extrinsics are estimated during SfM, while translational extrinsics come from the platform CAD model.

Known or intentional limitations

Reported constraints and characteristics to check before using the dataset.

Test-set camera poses are withheld for benchmark evaluation; validation poses are public.

Sparse LiDAR-derived depth maps are provided for training images rather than as dense, dataset-wide depth annotation.

Galaxy S9 images have irregular time delays caused by Android software synchronization and automatic exposure.

The environments include motion blur, low light, textureless and repetitive areas, reflective surfaces, changing displays, and substantial human occlusion.

Cameras

basler-camera-ring

Model
Basler acA2500-20gc
Setup
Mono
Count
6
Effective cameras
6
Modality
Rgb
Resolution
2592 x 2048
Rate
2.5 Hz
Shutter
Global
Lens type
Not Reported
HFOV
79.4 deg
VFOV
63 deg

galaxy-query-cameras

Model
Samsung Galaxy S9
Setup
Mono
Count
4
Effective cameras
4
Modality
Rgb
Resolution
2160 x 2880
Rate
1 Hz
Shutter
Rolling
Lens type
Not Reported
HFOV
Not reported
VFOV
Not reported

IMUs

Not reported

GNSS

Not reported

LiDAR

Velodyne VLP-16

Model or name
Velodyne VLP-16
Count
2
Dimensions
3
Rate
10 Hz
Channels
16
Range
Not reported

Additional sensors

Wheel encoder

Model
ams AS5047
Name
Magnetic rotary wheel encoder
Count
2
Resolution
1024 pulses per rotation

Annotations

Reported annotation availability and task support.

Availability
Available
Format
kapture; sparse LiDAR-derived depth maps for training images
Class count
Not reported
Semantic segmentation Not reported
Instance segmentation Not reported
Object detection Not reported
Optical flow Not reported
Depth ground truth Available

Citation

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

Citation key
Lee2021

BibTeX

@inproceedings{Lee2021,
  title = {Large-Scale {{Localization Datasets}} in {{Crowded Indoor Spaces}}},
  booktitle = {2021 {{IEEE}}/{{CVF Conference}} on {{Computer Vision}} and {{Pattern Recognition}} ({{CVPR}})},
  author = {Lee, Donghwan and Ryu, Soohyun and Yeon, Suyong and Lee, Yonghan and Kim, Deokhwa and Han, Cheolho and Cabon, Yohann and Weinzaepfel, Philippe and Guerin, Nicolas and Csurka, Gabriela and Humenberger, Martin},
  year = {2021},
  month = jun,
  pages = {3226--3235},
  publisher = {{IEEE}},
  address = {{Nashville, TN, USA}},
  doi = {10.1109/CVPR46437.2021.00324},
  urldate = {2023-10-04},
  isbn = {978-1-66544-509-2}
}

Dataset corrections

Notice a specific issue after reviewing the page?

Report an inaccuracy