Automated Recognition of Cold Water Coral Species through Object Detection

The seabed, extending beyond 200 meters in depth, represents the largest biome on the planet and hosts an extremely rich and complex biodiversity, yet it is threatened by various anthropogenic factors. Protecting and monitoring deep-sea ecosystems is essential to preserving marine biodiversity and achieving Good Environmental Status (GES) as mandated by the Marine Strategy Framework Directive (MSFD, 2008/56/EC). This thesis focuses on automating the recognition of certain Cold Water Coral (CWC) species, which form Vulnerable Marine Ecosystems (VMEs) in the Mediterranean Sea. The study leverages advanced technologies such as Object Detection and Deep Learning algorithms to create an automated system capable of accurately recognizing various coral species from visual data collected during underwater investigations. By integrating modern technological advancements with ecological monitoring, this work aims to improve the effectiveness of assessments of deep-sea ecosystems and contribute to the sustainable management of marine biodiversity. The findings highlight the potential of automated systems in environmental monitoring, paving the way for more frequent and comprehensive assessments of vulnerable marine habitats.