Self-supervised feature learning for rapid processing of marine imagery

The Postgraduate Research Scholarship in Marine Robotics has been developed to support an outstanding postgraduate research student in the Faculty of Engineering at the University of Sydney to undertake research in marine robotics. Quantitative estimates of marine environmental health are needed for scientific studies, design and management of protected areas, and regulatory compliance of industrial activity in the ocean. Australia is collecting seafloor images at increasing rates using underwater robots and other imaging systems, but expert annotations are not keeping up meaning that insights from this data are not being realised. This project will develop Machine Learning (ML) tools that leverage an already existing, large repository of labelled and unlabelled underwater imagery to train classifiers to derive quantitative estimates of species abundance and distributions for a variety of stakeholders. The project is relatively unique in that it provides opportunities to work alongside industry partners and academic researchers, as well as a pool of expert marine scientists who have a vested interest in using machine learning tools to solve their real-world image annotation challenges. These domain experts will provide curated, standardised training data for building applied ML models, and will also provide expert validation of these models.

We are seeking outstanding candidates to join our marine robotics team and to help develop the next generation of ML tools to work with seafloor imagery. You will have a background in machine learning or computer vision and a desire to see your work integrated into existing workflows with our partner organisations. You will also have the opportunity to participate in fieldwork designed to validate the developed tools and to collect new datasets to support your work using the latest marine robotic platforms.