AqUavplant Dataset: A High-Resolution Aquatic Plant Classification and Segmentation Image Dataset Using UAV

Area of Science:

• Ecology
• Remote Sensing
• Botany

Background:

• Aquatic vegetation is crucial for ecosystem stability but faces gradual decline.
• Effective conservation requires accurate monitoring and mapping of aquatic plant species.
• Unmanned Aerial Vehicles (UAVs) offer a viable solution for large-scale aquatic environment mapping.

Purpose of the Study:

• To develop the AqUavplant dataset for aquatic plant mapping and monitoring.
• To facilitate the development of machine learning models for automatic plant identification.
• To support biodiversity conservation and management of aquatic ecosystems.

Main Methods:

• Collected 197 high-resolution (4K) images of 31 aquatic plant species across nine sites in Bangladesh.
• Utilized a DJI Mavic 3 Pro drone with a Ground Sampling Distance (GSD) of 0.04-0.05 cm/px for detailed image acquisition.
• Created binary and multiclass semantic segmentation masks to complement the image dataset.

Main Results:

• The AqUavplant dataset provides comprehensive, high-resolution imagery of diverse aquatic flora.
• The dataset includes detailed segmentation masks essential for training machine learning algorithms.
• This resource enables precise mapping and monitoring of aquatic vegetation.

Conclusions:

• The AqUavplant dataset is a valuable resource for ecological research and conservation.
• It supports the detection of indigenous and invasive species, monitoring of plant health, and biodiversity assessment.
• The dataset aids in preventing aquatic plant extinction and preserving ecosystem balance.