List of Datasets

UAV

UAVScenes

semantic segmentation; localization; depth estimation; novel view synthesis

Annotation and pose extension of MARS-LVIG with paired aerial images and LiDAR clouds.

Metadata: Complete Metadata source documented Released 2025

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

Supplementary

Not reported

Quick facts

Vehicle type
UAV
Environment
Outdoor
Data origin
Real-world
Sensor count
2
Ground truth
Available
Calibration
Parameters Only
Annotations
Available

Sensor overview

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

Cameras

1 reported

IMUs

Not reported

GNSS

Not 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.
6DoF Structure from motion DJI Terra reconstruction Not reported Not reported Partial · 20 Reconstructed Sequences; Excludes Featureless GNSS01, Featureless GNSS02 and Featureless GNSS03 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.
Hybrid
Reference notes
Uses released camera images plus GNSS initializations; 20 sequences only; not an independent trajectory reference.
Other Other Projected Livox Avia clouds Not reported Not reported Partial 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
Per-image depth from corresponding input LiDAR; independent of camera depth predictor but not of LiDAR fusion.

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 interval1_CAM_LIDAR package includes calibration parameters; camera–LiDAR parameters are also documented in calibration_results.py. Raw calibration sessions are not established.

Known or intentional limitations

Reported constraints and characteristics to check before using the dataset.

An annotation and pose extension of MARS-LVIG, with 23 sequences in four outdoor areas; the parent collection is not an additional independent set of recordings.

Twenty sequences have reconstructed poses and annotated clouds. Featureless_GNSS01, Featureless_GNSS02 and Featureless_GNSS03 provide only dynamic-object image labels; no reconstructed poses or annotated clouds are supplied for them.

The eight reconstruction groups or maps are not the sequence count. Full-rate interval-1 downloads and interval-5 keyframe downloads have different sampling.

Camera poses reuse released imagery and GNSS initialization, and projected depth reuses input LiDAR. Neither is an independent reference for every sensor-fusion use case; reference accuracy is not established.

IMU and GNSS hardware in the parent collection does not establish independently released streams here. The encrypted DJI L1 raw clouds are outside the frame-wise annotation product.

The repository provides data-processing utilities, while the complete benchmark baseline code is still marked as under preparation in the checked documentation.

Cameras

Camera 1

Model
Hikvision
Setup
Mono
Count
1
Effective cameras
1
Modality
Rgb
Resolution
2448 x 2048
Rate
10 Hz
Shutter
Not Reported
Lens type
Not Reported
HFOV
Not reported
VFOV
Not reported
Notes
Exact model and shutter unspecified in this paper.

IMUs

Not reported

GNSS

Not reported

LiDAR

Livox Avia

Model or name
Livox Avia
Count
1
Dimensions
3
Rate
10 Hz
Channels
Not reported
Range
Not reported

Radar

Not reported

Additional sensors

Not reported

Annotations

Reported annotation availability and task support.

Availability
Available
Format
19-class image and point-cloud semantic labels; reconstructed semantic maps. Sequence coverage differs between the products.
Class count
19
Semantic segmentation Available
Instance segmentation Not reported
Object detection Not reported
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
wang2025uavscenes

BibTeX

@misc{wang2025uavscenes,
  author = {Wang, Sijie and Li, Siqi and Zhang, Yawei and Yu, Shangshu and Yuan, Shenghai and She, Rui and Guo, Quanjiang and Zheng, JinXuan and Howe, Ong Kang and Chandra, Leonrich and Srijeyan, Shrivarshann and Sivadas, Aditya and Aggarwal, Toshan and Liu, Heyuan and Zhang, Hongming and Chen, Chujie and Jiang, Junyu and Xie, Lihua and Tay, Wee Peng},
  title = {{UAVScenes}: A Multi-Modal Dataset for {UAVs}},
  year = {2025},
  eprint = {2507.22412},
  archivePrefix = {arXiv},
  primaryClass = {cs.CV},
  url = {https://arxiv.org/abs/2507.22412v1}
}

Dataset corrections

Notice a specific issue after reviewing the page?

Report an inaccuracy