Getting the most out of deep-sea images for ecological assessment

This project will investigate the sensitivity of seabed ecological metrics (e.g. organism density, biomass, diversity indices and community composition) derived from georeferenced imagery to survey design, image quality (e.g. resolution/footprint) and annotation strategies (e.g. human expert vs. state of the art machine learning). Labels generated for both habitat types spanning across multiple image frames and individual organisms within an image frame will be used in the analysis, and the work will establish best practises to optimise the use of image-derived data to monitor communities, both in the UK marine regulatory context and generalizable for further afield.