Segmentation of ocean ground elements using Mask-RCNN

In this work, we present a demonstration of the segmentation of ocean ground elements using the Mask R-CNN deep learning architecture. The method is applied to underwater imagery to accurately identify and segment various seafloor components such as rocks, sand patches, vegetation, and marine debris. Leveraging the power of convolutional neural networks and region-based proposals, Mask R-CNN enables precise object detection and pixel-wise segmentation, even in challenging underwater environments. This approach contributes to advancing marine research, environmental monitoring, and autonomous underwater navigation by providing detailed and reliable scene understanding. This work has been carried out by Ivan Carratala.