a case study from Doñana

View the PDF file for the paper entitled “bird song detector to improve the identification of birds through deep learning: a case study from DO \ ~ Nana, by Alba M \ ‘Arquez-Rodr \” IGUEZ and 5 other authors

PDF HTML (experimental) view

a summary:Negative audio monitoring is a major tool for maintaining biological diversity, but large quantities of non -supervisory sound, it generates current great challenges to extract meaningful information. Deep learning provides promising solutions. Birdnet, a widely used bird identification model, has been successful in many study systems, but is limited by domestic scale due to the biases in their training data, which focus on specific sites and targeted sounds instead of the entire audio tools. One of the main challenges is to identify the types of birds that many records either lack the target species or contain overlapping boxes, which complicates automatic identification. To address these problems, we have developed a multi -stage pipeline to determine the automatic definition of birds in the National Deniana Park (SW Spain), which is a wetland of high interest in conservation. We published Audiomoth records in three main habitats through nine sites and manually explained 461 minutes of sound, which led to 3749 sectors extending to 34 categories. We first apply bird song detector to insulate bird boxes using the spectrum -based image processing. Next, the species were classified using dedicated local models on a local scale. Bird song detector was applied before specifying species, as all models were performed better when analyzing the sectors that were discovered only. Specifically, a mixture of detector and modified birds outperformed the foundation line without discovery. This approach shows the effectiveness of combining bird songs with local classification models. These results highlight the need to adapt tools for general purposes for specific environmental challenges. Discovery of bird species automatically helps to track the health of this threatened ecosystem, given the sensitivity of birds for environmental change, and supports preservation planning to reduce the loss of biological diversity.