In a world where facial recognition technology is becoming commonplace, it is not just humans who are being identified. A groundbreaking study has successfully applied this technology to identify individual whales and dolphins in the wild across 24 species.

Led by Philip Patton, a PhD student at the University of Hawaiʻi at Mānoa's Hawaiʻi Institute of Marine Biology (HIMB), the research introduces a multispecies photo-identification model. This model was developed for a Kaggle competition organized by Happywhale.com. The competition challenged engineers to create a tool that could individually identify whales and dolphins using an algorithm.

Algorithm

The facial recognition model used in this study is based on a state-of-the-art method in human facial recognition known as the ArcFace classification head. It uses two such heads to jointly classify species and identities, allowing species to share information via shared weights within the network.

The algorithm can identify characteristics such as scarring, pigmentation and size on individual dolphins and whales. This accelerates the information-gathering process, allowing researchers to quickly estimate population size, population trends, and space use—all crucial factors for conserving Hawaiian whales and dolphins.

Ecologically, dolphins are very social, and this new tool provides a way to observe dolphin social behavior in a non-invasive way.