A global compilation of seafloor images, known as BenthicNet, has been created for deep learning applications. The compilation appears to include images from various locations, such as the Great Barrier Reef and the Clarion-Clipperton Zone, and is being used for tasks like identifying biota and annotating survey images. The dataset also includes expert-annotated images from Australian sea-floor surveys, which can be used to improve automated annotation and other applications.