Search

The increasing availability of seabed images has created new opportunities and challenges for monitoring and better understanding the spatial distribution of fauna and substrata. To date, however, deep-sea substratum classification relies mostly on visual interpretation, which is costly, time-consuming, and prone to human bias or error. Motivated by the success of convolutional neural networks in learning semantically rich representations directly from images, this work investigates the application of state-of-the-art network architectures, originally employed in the classification of non-seabed images, for the task of hydrothermal vent substrata image classification. In assessing their potential, we conduct a study on the generalization, complementarity and human interpretability aspects of those architectures. Specifically, we independently trained deep learning models with the selected architectures using images obtained from three distinct sites within the Lucky-Strike vent field and assessed the models' performances on-site as well as off-site. To investigate complementarity, we evaluated a classification decision committee (CDC) built as an ensemble of networks in which individual predictions were fused through a majority voting scheme. The experimental results demonstrated the suitability of the deep learning models for deep-sea substratum classification, attaining accuracies reaching up to 80% in terms of F1-score. Finally, by further investigating the classification uncertainty computed from the set of individual predictions of the CDC, we describe a semiautomatic framework for human annotation, which prescribes visual inspection of only the images with high uncertainty. Overall, the results demonstrated that high accuracy values of over 90% F1-score can be obtained with the framework, with a small amount of human intervention.

We set out to characterize the fine‐scale processes acting on interannual dynamics of deep‐sea vent fauna by using a novel approach involving a 5‐yr time series of 3D photogrammetry models acquired at the Eiffel Tower sulfide edifice (Lucky Strike vent field, Mid‐Atlantic Ridge). Consistently, with the overall stability of the vent edifice, total mussel cover did not undergo drastic changes, suggesting that they have been at a climax stage for at least 25 yr based on previous data. Successional patterns showed consistency over time, illustrating the dynamic equilibrium of the ecological system. In contrast, microbial mats significantly declined, possibly due to magmatic events. The remaining environmental variability consisted of decimeter‐scale displacement of vent outflows, resulting from their opening or closure or from the progressive accretion of sulfide material. As a result, vent mussels showed submeter variability in the immediate vicinity of vent exits, possibly by repositioning in response to that fine‐scale regime of change. As former studies were not able to quantify processes at submeter scales in complex settings, this pioneering work demonstrates the potential of 3D photogrammetry models for conducting long‐term monitoring in the deep sea. We observed that the ability of mussels to displace may enable them to cope with changing local conditions in a stable system. However, the long‐term stability of mussel assemblages questions their capacity to withstand large‐scale disturbances and may imply a low resilience of these “climax” communities. This suggests that they may be particularly vulnerable to the negative effects of mining activities in hydrothermal ecosystems.