Blue-Lined Rabbitfish form long-term pair bonds, making them a valuable indicator of reef health under climate stress. If these partnerships change, weaken, or disappear, they may offer early clues about how warming oceans are reshaping reef ecosystems. But traditional monitoring relies on invasive physical tagging that can stress the very fish researchers are trying to study, while manual photo-identification across thousands of dive images is prohibitively slow.

ReefPairs replaces the tag with the camera. Developed successfully as a prototype, the platform offers a non-invasive, AI-assisted re-identification web application. It uses a custom detection model to identify rabbitfish in underwater imagery with up to ~98% precision, and a visual re-identification module that suggests the top identity matches with 79.7% accuracy. Researchers stay in the loop throughout: reviewing detections, using visual enhancement tools to confirm identity, and manually linking observed pair bonds.

Ultimately, ReefPairs moves marine monitoring from simple species presence towards individual-level and pair-level understanding. By linking who is seen, where, when, and with whom, the platform creates a scalable foundation for tracking reef fish relationships, site fidelity, habitat use, and behavioural responses to a warming ocean.