ESC

Software

Our analysis pipelines are released openly. Everything below is built to run over large image collections without manual annotation of every frame.

AI-SCW — Automated and Integrated Seafloor Classification Workflow

Authors: B. Mbani, T. Schoening, J. Greinert

An end-to-end workflow for classifying seafloor imagery into habitat classes without hand-labelling every image. Developed on manganese-nodule covered seabed in the Central Pacific and published in Scientific Reports (2022). It is the classification backbone we are now adapting to mangrove and coastal habitats.

FaunD-Fast — Megabenthic Fauna Detector

Authors: B. Mbani, V. Buck, J. Greinert

A Faster R-CNN detector for locating and counting megabenthic fauna in optical seafloor images, trained and evaluated on imagery from the Clarion–Clipperton Zone. Released via Zenodo and GEOMAR, and the direct ancestor of the computer-vision component of Blue-BII.

Analysis-ready seafloor imagery, SO268 cruises

Authors: B. Mbani, J. Greinert

An open, colour-corrected and georeferenced image dataset collected along OFOS profiles during RV SONNE cruises SO268/1 and SO268/2, published in Scientific Data (2023) and archived at PANGAEA.