(2025). floating marine litter dataset. SEANOE. https://doi.org/10.17882/106148
Copy this text
floating marine litter dataset
The Floating Marine Litter (FML) Dataset comprises approximately 5500 floating marine litter optical images and 14000 bounding box annotations, tailored for the object detection task. The images were taken from a first-person perspective using a camera mounted on an Unmanned Surface Vehicle (USV). The images were collected in diverse lighting conditions and camera setups across inland and coastal water bodies.
Disciplines
Environment
Parameters
Devices
Keywords
Floating Marine Litter, Object Detection, First-person View, Unmanned Surface Vehicles (USV)
Location
45.788116N, 45.771834S, 15.940475E, 15.903397W
43.953778N, 43.927078S, 15.45639E, 15.415192W
Devices
The optical images were taken using a Hikvision PTZ Camera DS-2DE2A404IW-DE3, a compact, high-performance pan-tilt-zoom (PTZ) camera. The sensor specifications are:
- Resolution: The camera provides a high-definition resolution of 1920×1080 pixels;
- PTZ Functionality: it features two rotational degrees of freedom, with a pan range of 0° to 330° and a tilt range from -5° to 90°. Additionally, it incorporates a 4× optical zoom, supporting a focal length range of 2.8 mm to 12 mm;
- Shutter Speed: the shutter speed is adjustable both automatically, based on ambient lighting condition, and manually, with a configurable range from 1/30000 s to 1/25 s;
- Image Enhancement: the built-in software allows for customization of brightness, contrast, saturation, and sharpness. Furthermore, the Wide Dynamic Range (WDR) functionality improves image quality in scenes with high contrast between bright and dark areas;
- Infrared Capability: an integrated infrared sensor enables night vision, facilitating image acquisition in low-light conditions;
Environmental Robustness: the camera is IP66-rated, preventing water from entering the camera case.
The camera’s versatility enabled the acquisition of images under varying environmental and lighting conditions to collect a rich dataset.
Data
| File | Size | Format | Processing | Access | |
|---|---|---|---|---|---|
data | 2 Go | IMAGE | Raw data |
How to cite
