We are excited to share that our new paper, "Whale Vision: A tool for identifying sperm whales and other cetaceans by their flank or fluke", has just been published.
Our research introduces machine learning methods to automate the re-identification of individual sperm whales from photos, and the identification of previously unseen whales, a process that typically requires labour-intensive manual curation. We developed Residual Neural Network models capable of identifying sperm whales from both fluke and, for the first time, flank images, achieving high accuracy on the Oceanomare Delphis dataset. Additionally, we have demonstrated that our approach is effective for other dolphin species, whether using the original sperm whale networks or by retraining the networks for greater accuracy.
Our user-friendly application utilises these networks to facilitate faster and more scalable whale identification and re-identification, and is available at github.com/whale-vision.