Deep-Sea Fauna Segmentation: A Comparative Analysis of Convolutional and Vision Transformer Architectures at Lucky Strike Vent Field

The impact of deep-sea mining will depend critically on the ability of larval dispersal of hydrothermal mollusks to connect and replenish natural populations. However, assessing connectivity is extremely challenging, especially in the deep sea. Here, we investigate the potential of using the chemical composition of larval shells to discriminate larval origins between multiple hydrothermal sites in the southwest Pacific. Our results confirm that this method can be applied with high accuracy.