Marine surface
WaterScenes
Multi-task water-surface radar-camera perception
USV recordings provide 54120 labeled image/radar/GPS/IMU sets across inland waterways, lighting and weather conditions.
Access links
Primary places to inspect, download, read about, or reproduce the dataset.
Quick facts
- Vehicle type
- Marine surface
- Environment
- Outdoor
- Data origin
- Real-world
- Sensor count
- 5
- Ground truth
- Not reported
- 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.
Not reported
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 manifest lists calib intrinsics/extrinsics for projection. Complete raw sessions and independent alignment accuracy remain unknown.
Known or intentional limitations
Reported constraints and characteristics to check before using the dataset.
The 54,120 matched multimodal sets are keyframes, not independent recording sequences. Free-space, waterline and object/radar annotation products have distinct counts and taxonomies.
Camera keyframes are 2 Hz from 30 Hz capture; radar associations and three/five-frame accumulation differ from the sensor 15 Hz rate. Full-rate GPS/IMU/radar delivery is not established.
IMU CSV documents angular/linear velocity and magnetic fields, without proving native accelerometer/gyro sampling or raw pressure delivery. GPS ground speed/heading are associated inputs, not a surveyed reference trajectory.
Fifty-millisecond nearest-time tolerance, water reflections, radar dropout, distant objects and lens waterdrops affect fusion. The later WaterScenes_Vision release is a separate packaging update, not proof of more original multimodal sequences.
Cameras
Camera 1
- Model
- Sony IMX317 CMOS sensor
- Setup
- Mono
- Count
- 1
- Effective cameras
- 1
- Modality
- Rgb
- Resolution
- 1920 x 1080
- Rate
- 2 Hz
- Shutter
- Not Reported
- Lens type
- Not Reported
- HFOV
- Not reported
- VFOV
- Not reported
- Notes
- Annotated release keyframes are 2 Hz; camera capture is 30 Hz. Mounted observation network camera is not established as a released stream.
IMUs
WaterScenes AHRS
- Model
- Not reported
- Count
- 1
- Accel / gyro
- Not reported
GNSS
GNSS 1
- Model or name
- Not reported
- Count
- 1
- Rate
- Not reported
- GNSS type
- STANDARD
- Position output
- Available
- Velocity output
- Available
- Heading output
- Available
LiDAR
Not reported
Radar
Oculii EAGLE
- Count
- 1
- Released data
- Point detections / point clouds
- Frame rate
- 2 Hz
- Doppler velocity
- Available
- Elevation angle
- Available
- Configured range
- Not reported
- Field of view
- 110° H × 45° V (per radar)
- Carrier frequency
- 77 GHz
- File format
- CSV point clouds
- Notes
- Released products associate radar point sets to 2 Hz image keyframes, with separate three/five-frame accumulations. The sensor captures at 15 Hz; these are different cadences, not a universal raw 15 Hz release.
Additional sensors
Magnetometer
- Name
- AHRS magnetic-field samples
- Count
- 1
- Notes
- Official matched IMU CSV lists magnetic_field_x/y/z from the same AHRS package. Native cadence and chip are unknown; no separate extra AHRS/IMU package is implied.
Annotations
Reported annotation availability and task support.
- Availability
- Available
- Format
- 2D boxes, image polygons, waterline polylines and radar point class/instance IDs; seven object classes plus task-specific free-space/clutter
- Class count
- Not reported
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.
@ARTICLE{yao2024waterscenes,
author={Yao, Shanliang and Guan, Runwei and Wu, Zhaodong and Ni, Yi and Huang, Zile and Liu, Ryan Wen and Yue, Yong and Ding, Weiping and Lim, Eng Gee and Seo, Hyungjoon and Man, Ka Lok and Ma, Jieming and Zhu, Xiaohui and Yue, Yutao},
journal={IEEE Transactions on Intelligent Transportation Systems},
title={WaterScenes: A Multi-Task 4D Radar-Camera Fusion Dataset and Benchmarks for Autonomous Driving on Water Surfaces},
year={2024},
volume={25},
number={11},
pages={16584-16598},
doi={10.1109/TITS.2024.3415772}}Dataset corrections
Notice a specific issue after reviewing the page?
Report an inaccuracy