PLAN-SUBSIM delivers first object detection models for benthic fauna and habitat builders

Report summary

The increasing access to autonomously operated technologies offer vast opportunities to sample large volumes of biological data. However, these technologies also impose novel demands on ecologists to apply efficient tools for data management and processing, which are publicly available and easy to use. In order to address these needs Ocean Data Factory Sweden (ODF) and Swedish Biodiversity Data Infrastructure (SBDI) develop a national platform for analysis of subsea images (PLAN-SUBSIM). The system is named Koster Seafloor Observatory (KSO) and combines several essential functions for analysing large volumes of subsea image data for marine ecological research in Sweden. The system has now finished its 3^rd developmental cycle which produced open-source services to upload, explore, and manage image subsea classifications, as well as train and test object detection models.

The third cycle also included a number of scientific collaborations with Geological Survey of Sweden (SGU) and Medins AB on sea bed monitoring in the Kattegat and Baltic sea. As a result, we can now offer a first suite of object detection models for 5 key ecological species, including sea pens (Pennatula phosphorea, Virgularia mirabilis), the Deeplet sea anemone (Bolocera tuediae), the cold-water coral (Desmophllum pertusa), and blue mussels (Mytilus spp.). The models are verified, well documented, and ready for use in ecological research and seafloor monitoring in Swedish coastal waters. An example of model performance can be found here, while our model zoo is accessible here.

Image credits: Koster Seafloor Observatory.