Individual identification of endangered amphibians using deep learning and smartphone images: case study of the Japanese giant salamander (Andrias japonicus)

Abstract

Abstract Information obtained via individual identification is invaluable for ecology and conservation. Physical tags, such as PIT tags and GPS, have been used for individual identification; however, their impact on animal behavior and survival rates is unclear and the tags may become lost. Although non-invasive methods that do not affect the target species (such as manual photoidentification) are available, these techniques utilize stripes and spots that are unique to the individual, which requires training, and applying them to large datasets is challenging. Many studies that have applied deep learning for identification have focused on species-level identification; however, few have addressed individual-level identification. In this study, we developed an image-based identification method with deep learning using the head spot of the Japanese giant salamander (Andrias japonicus), an endemic and endangered species in Japan. We trained and evaluated the dataset collected over two days from 11 individuals in captivity, including 7,075 images from the smartphone camera. Photographing was conducted three times a day at approximately 11:00 (morning), 15:00 (evening), and 18:00 (afternoon). As a result, individual identification by EfficientNet-V2 achieved 99.86% accuracy, 0.99 Kappa coefficient, and 0.99 F1 score. Performance was lower in the evening than in the morning or afternoon when trained and evaluated at each photographing time. This method does not require direct contact with the target species, and the effect on the animals is minimal; moreover, individual-level information can be obtained under natural conditions. In the future, smartphone images can be applied to citizen science surveys and individual-level big data collection, which is difficult to perform using current methods.