Ecosystem Monitoring using Deep Learning

The project will develop methods to use acoustic data for the identification of animals in the wild and in controlled agricultural settings. The initial objective is to capture data with low-cost devices that can be uploaded to a cloud-based service, in order to perform species identification in real-time using deep learning networks. Beyond species classification and occurrence counts, the project will investigate the possibility to use advanced machine learning techniques for the identification of individual-level and colony-level behaviours and health status, which is particularly important for automatic hive monitoring in precision apiculture. Data fusion with visual information may be considered as an extension to extend the range of detection capabilities.