Crown-of-thorns starfish (COTS) are found throughout the Indo-Pacific region and feed on coral polyps. Normally, their numbers are too small to harm the ecosystem, but when outbreaks increase in frequency they cause significant damage by consuming the polyps faster than they can grow.

Healthy reefs take about 10 to 20 years to recover from COTS outbreaks, and degraded reefs are less likely to recover, resulting in irreversible damage.

COTS outbreaks can be controlled by injecting starfish with bile salts or removing populations from the water.

The problem is that surveying reefs for outbreaks by towing a snorkeler behind a boat for visual identification is time-consuming, labour-intensive, and not particularly accurate. So CSIRO developed an edge ML platform based on the NVIDIA Jetson AGX Xavier that can detect COTS from live camera feeds in near real time at more than 10fps.

The model can cope with varying ocean conditions including lighting, visibility, depth, viewpoint, coral habitat, and the number of COTS present.

1080p images are processed at 11fps, and 720p images at 22fps with only a slight reduction in accuracy.

The project, is part of Google's Digital Future Initiative with CSIRO, shows how machine learning and AI technology can be applied to the large-scale surveillance of ocean habitats.

The work is open source through the crown-of-thorns starfish detection pipeline on GitHub or on Google Colab.