2025-03-14 京都大学
西表島では、外来種であるシロアゴガエルとオオヒキガエルが近隣の石垣島からたびたび侵入している。本研究では侵入の早期発見に役立てるため、これら2種の鳴き声を検出するAIを開発した。(撮影:福山伊吹、木村楓)
<関連情報>
- https://www.kyoto-u.ac.jp/ja/research-news/2025-03-14-2
- https://www.kyoto-u.ac.jp/sites/default/files/2025-03/web_2503_Kimura-bdd15980985367dd6fa9212f109fa09e.pdf
- https://link.springer.com/article/10.1007/s10530-025-03553-0
侵入前線にある島におけるディープラーニングに基づく侵入外来カエルPolypedates leucomystaxとRhinella marinaの検出 Deep learning-based detector of invasive alien frogs, Polypedates leucomystax and Rhinella marina, on an island at invasion front
Kaede Kimura,Ibuki Fukuyama & Kinji Fukuyama
Biological Invasions Published:04 March 2025
DOI:https://doi.org/10.1007/s10530-025-03553-0
Abstract
Early detection is crucial for managing invasive alien species at invasion fronts. Automated bioacoustic monitoring using deep learning can support this by detecting species calls. However, its application for invasive alien species remains limited. Additionally, the lack of local training data in pre-establishment areas, where invasive species are absent or rare, can limit model performance due to domain shifts. This study aimed to develop and evaluate a deep learning-based detector for Southeast Asian treefrog (Polypedates leucomystax) and cane toad (Rhinella marina) on Iriomote Island, a Natural World Heritage Site located 30 km from established populations of these alien species on the nearby Ishigaki Island. We trained the BirdNET model with acoustic data collected on Ishigaki Island, where these species were common, as well as native frog calls from Iriomote Island. Model performance was evaluated using (1) wild sounds on Ishigaki Island, and (2) sounds obtained by playing back the alien species call on Iriomote Island. Model precision and recall for wild sounds were 0.972 and 0.906 for P. leucomystax, and 0.899 and 0.705 for R. marina, respectively. Playback survey dates were identifiable from the high number of detections when the detection thresholds were properly adjusted, except on a night with dense choruses of a native frog species, Microhyla kuramotoi. These results suggest that data from Ishigaki Island enabled training a model with adequate generalization across invasion front.
