Translate 
Abstract
Seagrass meadows are essential to the health of coastal ecosystems. They support carbon storage, provide habitats for marine species, and help stabilize coastlines. Monitoring underwater seagrass is vital for understanding the conditions of the ecosystem. Researchers have been interested in identifying and classifying underwater seagrasses. However, traditional monitoring methods can be labor-intensive and costly, especially in complex underwater environments. Deep learning approaches have made significant progress in digital image processing, particularly in object recognition and classification, and are among the most popular computer vision tools. The collection of digital images for monitoring underwater habitats, such as seagrass meadows, has increased significantly as recent progress in imaging technology has made it easier to collect high-resolution data. The increase in imagery data has in turn created a demand for automated detection and classification using deep neural network-based techniques. This study reviews the current deep-learning techniques used for monitoring and classification of the seagrass. It discusses the key methodologies, datasets, and progress in this area. This study not only examines the well-known challenges such as limited availability of data but provides a novel, structured taxonomy of deep learning techniques tailored for the monitoring of seagrass, highlighting their unique advantages and limitations within diverse marine environments. By synthesizing findings across various data sources and model architectures, we offer critical insights into the selection of context-aware algorithms and identify key research gaps, an essential step for advancing the reliability and applicability of AI-driven seagrass conservation efforts.
Highlights
The ocean covers approximately 71% of the earth’s surface (Hawthorne 2019), and seagrass meadows are vital for the environmental resilience of coastal regions
This paper addresses the significant challenges posed by ecological changes driven by climate change and human activities, which makes the classification of seagrass increasingly complex and crucial for conservation efforts
This research survey effectively demonstrated the application of advanced deep-learning approaches for the accurate classification and mapping of seagrass habitats
Summary
The ocean covers approximately 71% of the earth’s surface (Hawthorne 2019), and seagrass meadows are vital for the environmental resilience of coastal regions. Seagrasses are key contributors to coastal health; they produce oxygen and are a significant carbon sink (Fig. 1) These ecosystems face growing threats from human activities, pollution, and climate change, which lead to the loss of habitats and reduced availability of light for photosynthesis (Beca-Carretero et al 2024). The monitoring of seagrass is essential for effective marine ecological management (Borum et al 2004) It helps track ecosystems and health, supports conservation efforts, and enables the early detection of environmental changes, contributing to the preservation of these valuable underwater habitats. Background removal isolates the subject, such as seagrass, by removing extraneous elements, such as water or sand, from the image. This is accomplished by thresholding or machine learning methodologies such as the Normalized Difference Vegetation Index (NDVI), which focuses on vegetation features. Zero-short CLIP (Jeon et al 2021) UCL SeaFearts + SeaCLIP (Raine et al 2024) EfficientNet-B5 (Noman et al 2021b) VGG-16 with 2-Layer + Drop (Raine et al 2020) SimCLR (Jeon et al 2021)
Sign-in/Register to view highlights & summary 
Published Version
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
