Handheld
WildScenes
2D/3D semantic segmentation
Handheld forest traversals with coherent RGB/lidar semantics, SLAM poses and label uncertainty histograms.
Access links
Primary places to inspect, download, read about, or reproduce the dataset.
Quick facts
- Vehicle type
- Handheld
- 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.
Ground truth
Reference scope, method, rate, coverage, and provenance.
| Scope | Method | Reference system | Count | Rate | Coverage |
Reference provenance
|
|---|---|---|---|---|---|---|
| 6DoF | LiDAR SLAM | Wildcat with GPS/loop-closure bundle adjustment | Not reported | Not reported | Full |
Reference provenance
Hybrid
|
|
||||||
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
Camera parameters and sensor transforms accompany the release. Full raw calibration recordings and empirical alignment accuracy are not established.
Known or intentional limitations
Reported constraints and characteristics to check before using the dataset.
Five traversals cover two Australian forest areas and 21.28 km. Released image sampling follows distance/heading changes, so capture-camera and motor frequencies are not output frame rates.
Eighteen raw classes differ from fifteen 2D and twelve 3D evaluated classes. Label projection/visibility filtering and multi-view disagreement affect the 3D semantics; histograms retain uncertainty.
Wildcat poses combine shared LiDAR/inertial processing with GPS/loop closures. Independent pose accuracy and standalone raw IMU/GPS streams remain unestablished; later full clouds differ from labelled submaps.
Cameras
Camera 1
- Model
- e-CAM130A CUXVR
- Setup
- Mono
- Count
- 1
- Effective cameras
- 1
- Modality
- Rgb
- Resolution
- 2016 x 1512
- Rate
- Not reported
- Shutter
- Not Reported
- Lens type
- Not Reported
- HFOV
- Not reported
- VFOV
- Not reported
- Notes
- Front-camera annotated release only. Capture hardware runs 15 Hz, but release is trajectory-sampled every 5 m or accumulated 5° heading change.
IMUs
Not reported
GNSS
Not reported
LiDAR
Velodyne VLP-16
- Model or name
- Velodyne VLP-16
- Count
- 1
- Dimensions
- 3
- Rate
- Not reported
- Channels
- 16
- Range
- Not reported
Radar
Not reported
Additional sensors
Not reported
Annotations
Reported annotation availability and task support.
- Availability
- Available
- Format
- Indexed/palette PNG, SemanticKITTI-style .label files and per-point label-histogram CSVs. Manual RGB labels propagate to 3D; histogram uncertainty is not a second geometric reference.
- Class count
- Not reported
Semantic segmentation
Available
Instance segmentation
Not available
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.
@misc{vidanapathirana2023wildscenes,
author = {{Kavisha Vidanapathirana} and {Joshua Knights} and {Stephen Hausler} and {Mark Cox} and {Milad Ramezani} and {Jason Jooste} and {Ethan Griffiths} and {Shaheer Mohamed} and {Sridha Sridharan} and {Clinton Fookes} and {Peyman Moghadam}},
title = {WildScenes: A Benchmark for 2D and 3D Semantic Segmentation in Large-scale Natural Environments},
year = {2023},
eprint = {2312.15364},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2312.15364v2}
}Dataset corrections
Notice a specific issue after reviewing the page?
Report an inaccuracy