The Australian Institute of Marine Science (AIMS) and CSIRO, Australia’s national science agency, are working together to protect the Great Barrier Reef from the Crown-of-Thorns Starfish (COTS), a major predator that can cause widespread coral loss during outbreaks. Leveraging artificial intelligence (AI) and a specialised towed underwater imaging platform, the project is developing real-time machine learning models to detect and classify COTS, coral cover, and the scars left on corals by starfish feeding.

This collaboration supports the work of the COTS Control Teams and Queensland Parks and Wildlife Service (QPWS), who coordinate efforts to remove starfish and reduce damage to the reef. FileSender, AARNet’s large file transfer service, plays an essential role by enabling the fast, secure sharing of massive datasets critical to building and refining these models.

Sharing terabytes of imagery

Each field trip using the underwater imaging platform generates up to a quarter of a million high-resolution images, equating to many terabytes (TB) of data. AIMS collects these images using a dual-camera system mounted on the towed platform, capturing the benthic environment at four frames per second during surveys lasting up to two hours.

To develop accurate AI models, these large image collections need to be transferred to CSIRO, where domain experts annotate the data and retrain machine learning models. The annotated images and updated model parameters then flow back to AIMS, ensuring that field systems are equipped with the latest detection capabilities to support COTS control operations.

From hard drives to FileSender

Before adopting FileSender, AIMS relied on physically shipping hard drives via post or using commercial cloud storage services. These methods were slow, cumbersome, and impractical for managing the scale of data involved.

According to Scott Bainbridge, Marine Imaging Specialist at AIMS, FileSender has transformed the workflow. After each survey trip, AIMS teams download the images from the survey platforms, create local backups, compress the data into survey folders (each around 1 TB), and use FileSender to transfer them securely to CSIRO. Once processed, CSIRO returns the updated models via FileSender, ready for deployment in the field. These outputs also contribute to broader reef health monitoring systems like ReefCloud.

The success of these large data transfers depends not only on FileSender but also on AARNet’s high-speed research and education network. The network provides the reliable, high-capacity infrastructure that enables terabytes of data to move efficiently between AIMS and CSIRO.

“FileSender has made a huge difference to how we work and streamlined the way we collaborate with CSIRO,” said Scott. “Transferring such a large amount of data used to take days. It can now be done in a fraction of the time, without storage limits to manage and with confidence that the data is secure.”

In this vital collaboration, AARNet’s FileSender has proven to be an essential tool to enable fast, secure, and reliable data sharing that helps AIMS and CSIRO focus on their mission to safeguard the Great Barrier Reef.

Photo: Researchers launching an underwater imaging platform to survey the Great Barrier Reef. Credit: AIMS | Marie-Roman.