Deep Insights with Exploring Plankton Communities through Artificial Neural Networks

Abstract

The separation and categorization of marine species are of great significance for determining which interact differently among themselves and whether such interaction influences the maintenance of biodiversity in the marine ecosystem. In this study, we present a comparative analysis of various artificial intelligence (AI) models applied to the classification of images of two distinct types of polychaeta: Polychaeta species Type A and Type F. Applying a dataset holding annotated images of polychaeta type subdivision, we experience the effective AI architectures, e.g., CNNs and others, in the process of precise identification of these species. By means of methodology, we will process the image dataset and then train and evaluate different AI models on the sorted photos that were selected. The model metrics are evaluated, and the evaluation is based on the accuracy and the epoch for each model type. Moreover, through the study of the models’ decisions’ interpretability, we draw a darker picture of the essence of the classification processes. The result of our study has been decisive; it has indicated the advantages and disadvantages of each AI model when inspecting Type A and Type F polychaeta. Moreover, we consider the bearing of the result obtained on marine ecology research and highlight the prospect of AI-based image classification techniques in providing remote and real-time monitoring, subduction, and study of underwater biodiversity. This research moves forward with the efforts of artificial intelligence techniques in ecological studies, which further show how the combination of different disciplines can be used to solve problems of marine science and conservation.