Convolutional neural networks for hydrothermal vents substratum classification: An introspective study

Pedro Juan Soto Vega, Panagiotis Papadakis, Marjolaine Matabos, Loïc Van Audenhaege, Annah Ramiere, Jozée Sarrazin, Gilson Alexandre Ostwald Pedro da Costa

Abstract: Highlights•We evaluate six state-of-the-art convolutional neural networks for classifying images of hydrothermal vent sea floors.•We analyze uncertainty in a set of deep learning models for semi-automatic operational feasibility.•We conduct a visual interpretability analysis to explain model decision-making in the application.