Wildlife detection is a crucial aspect of the monitoring process in Tidmarsh. It provides information regarding bio-diversity and its evolution over the restoration process. Tidzam analyzes the audio streams provided by the 32 microphones in real-time in order to detect bird and other wildlife calls, as well as environmental phenomenoma such as rain, wind, and aircraft.
Several deep-networks are used in sequence, beginning with a general classification of environmental sounds as weather, birds, peepers, etc, and followed by expert classifiers for bird specimen identification. The detection events are stored in a database in order to visualize and analyze their spatial distribution over time. This provides valuable data about animal settlements, migrations and biodiversity, particularly during the night, when it would be difficult for human experts to identify birds. The system also flags sounds it doesn't recognize, such as previously unobserved birdcalls, in order to extract a recording which can be identified by a human expert.
Try it live: TidZam.