Dataset of High-Resolution Aerial Images for Intertidal Macroalgae.

Area of Science:

• Marine Biology
• Ecology
• Remote Sensing

Background:

• Macroalgae are crucial for benthic communities and ecosystem health, acting as stress indicators.
• Manual monitoring is labor-intensive and impractical for large-scale assessments.
• Remote sensing, especially using Unmanned Aerial Vehicles (UAVs), offers a more efficient alternative.

Purpose of the Study:

• To address the lack of specialized macroalgae image datasets for machine learning.
• To provide a high-resolution dataset of intertidal macroalgae from the NE Atlantic.
• To facilitate the development of machine learning models for macroalgae classification and semantic segmentation.

Main Methods:

• Collected UAV and in situ RGB imagery of 33 intertidal macroalgae species.
• Created three sub-datasets: photoquadrats (507 images), orthoimages (7,954 polygons), and individual labels (7,685 images).
• Trained a Convolutional Neural Network (CNN) on the developed dataset.

Main Results:

• Achieved a test accuracy of 86.72% for classifying 11 macroalgae classes using a CNN.
• Demonstrated the feasibility of using the created dataset for machine learning applications.
• Established a valuable resource for macroalgae research and monitoring.

Conclusions:

• The developed open-access dataset supports advanced machine learning for macroalgae identification.
• UAV-based remote sensing combined with machine learning is effective for monitoring intertidal macroalgae.
• This resource will advance ecological studies and conservation efforts for macroalgae ecosystems.