Sound ID is a new feature in the Merlin Bird ID app that allows users to listen to the birds around them and see live predictions of which species are singing. The app currently identifies 458 bird species in the U.S. and Canada based on their sounds, with more species and regions to come. Sound ID works by converting audio recordings into spectrogram images, which are then analyzed by a deep convolutional neural network. The model is trained using audio data that includes the precise moments when each bird is vocalizing, resulting in more accurate predictions. The development process involves collaboration between sound ID experts, machine learning team members, and field testers. Merlin draws inspiration from previous projects and uses fine-grained labels to improve performance. Future articles will delve into design decisions and upcoming developments for Sound ID.
https://www.macaulaylibrary.org/2021/06/22/behind-the-scenes-of-sound-id-in-merlin/