Seagrass Detection and Classification from Uderwater Images Using Deep Transfer Learning

Introduction: Seagrass is an important marine ecosystem that provides habitat for many species of fish, crustaceans, and other marine animals. However, a number of issues, such as pollution, overfishing, and climate change, pose a threat to seagrass beds and may cause the environments that support them to deteriorate. One of the main challenges in the conservation of seagrass ecosystems is the accurate classification of seagrass species using computer vision techniques. Convolutional Neural Networks (CNNs), one type of machine learning approach, have garnered more attention lately for the purpose of classifying seagrass species. However, there are several challenges associated with using CNNs for seagrass classification. One of the main challenges is the lack of high-quality, labeled data. Another challenge is the variability in seagrass appearance and its complex growth. Therefore, it is obvious to keep track and identify the types of seagrass present in the coastal region. The proposed image classifier with CNN based on VGG-16 network is effectively used to classify the seagrass into five categories like Syringodium Isoetifolium, Halophila Ovalis, Enhalus Acoroides and Cymodocea Serrulata. By comparing the suggested categorization approach with the most advanced techniques, an overall accuracy of 90.50% is achieved.