Comparative Analysis of CNN Models for Semantic Segmentation of Posidonia Oceanica Meadows

Posidonia oceanica (PO) meadows are crucial underwater ecosystems that provide significant environmental benefits, including carbon sequestration and habitat for marine life. Accurate monitoring and mapping of these meadows are essential for their conservation but challenging due to the complexities of the subsea environment. This paper presents a comparative analysis of various Convolutional Neural Network (CNN) models for Semantic Segmentation of PO in underwater optical imagery. The study evaluates several state-of-the-art CNN architectures, including U-Net, Feature Pyramid Network (FPN), Pyramid Attention Network (PANet), Pyramid Scene Parsing Network (PSPNet), and DeepLabV3+. These models are trained and tested on multiple datasets to assess their performance in identifying and segmenting PO. The analysis considers relevant indices such as Average Precision (AP) and Area Under the Curve (AUC) to provide a comprehensive quantitative evaluation. This work aims to enhance the automation of marine ecosystem monitoring, contributing to the preservation and management of these vital underwater habitats.