AIMS had developed software called ‘BenthoBox’ to automatically process images of benthic (seabed) habitat taken during ecological surveys, a task traditionally done by marine ecologists. The software uses state-of-the-art deep learning algorithms ─ similar to those that drive the likes of Google and Facebook recognition systems ─ to recognise ‘tagged’ seabed features such as sand, algae, sponges, and corals.
BenthoBox is enhancing ecological reporting by analysing more imagery at a much faster rate and at a finer scale than previously possible. It identifies the contents of hundreds of thousands of sea floor images, enabling biologists and ecologists to better understand and determine indicators of seabed health, including changes in the density of flora and fauna.
This breakthrough technology is being used with imagery taken on the Great Barrier Reef as part of the AIMS Long Term Monitoring program, and being used to screen imagery captured by towed camera systems in deeper waters off north Western Australia.
It represents a step-change in the assessment and monitoring of marine flora and fauna across multiple geographical scales and will support ecosystem-based management decision making and monitoring programs into the future.
BenthoBox is an automated image classification program specifically designed to learn and accurately detect benthic (or sea floor) features for thousands of images.