Machine Learning Aids Sea Floor Classification

Scientists can now automatically process and classify data about large areas of the sea floor with a new machine-learning method developed by the Australian Institute of Marine Science (AIMS).

The new method, RapidBenthos, has the capability to extract an unprecedented level of data from photomosaics of coral reefs.

It automates analysis of mosaics stitched together from thousands of seafloor images, saving scientists about 60 hours of manual analysis per mosaic.

It can be used for monitoring coral reefs, coral bleaching and seagrass meadows, providing timely information to inform management decisions.

Lead author, AIMS machine learning/AI engineer Tiny Remmers, said RapidBenthos allowed AIMS to more sustainably scale up and process more seafloor imagery with much less work and cost.

AIMS ecosystem modeller Dr Renata Ferrari, said RapidBenthos was possible thanks to years of work from a multidisciplinary team of researchers whose work towards capacity building in ecological and ecosystem monitoring supports many projects across AIMS and in the world.