Search papers, labs, and topics across Lattice.
This study reformulates bird vocalization detection as an object detection task using YOLO11 models to accurately localize bird calls within dense tropical soundscapes. By introducing a novel evaluation metric, Intersection over Minimum (IoMin), the researchers address the limitations of traditional metrics in handling ambiguous acoustic boundaries. The best-performing YOLO model achieved an impressive 81.8% IoMin@50 F1-score on in-distribution soundscapes, nearly doubling the baseline performance, and also outperformed the baseline on out-of-distribution recordings from Hawaii, highlighting the effectiveness of this approach in complex environments.
Object detection frameworks can nearly double the accuracy of bird call localization in noisy environments, transforming wildlife monitoring practices.
Passive acoustic monitoring enables large-scale observation of wildlife, but most bioacoustic classifiers only predict species presence in a time window without localizing vocalizations precisely in time or frequency, limiting downstream analyses. We formulate bird vocalization detection as an object detection task on spectrograms and train YOLO11 models to localize bird calls in dense tropical soundscapes from Singapore. We additionally introduce an open-source browser-based annotation tool and propose Intersection over Minimum (IoMin), an evaluation metric that better handles ambiguous acoustic boundaries than standard IoU and is better suited to the problem at hand. The best YOLO model nearly doubles baseline performance on in-distribution soundscapes from Singapore (81.8% vs. 42.1% IoMin@50 F1-score) while still outperforming the baseline on unseen out-of-distribution recordings from Hawaii (58.6% vs. 48.6%). These results suggest that object detection frameworks are a promising approach to time-frequency localization of animal vocalizations in complex soundscapes.
NUS