A network of wildlife sanctuaries has led to the reintroduction of many of New Zealand's native bird species, and now machine learning software could be an important tool in monitoring the survival of these birds once they leave these protected areas.
Sanctuaries have been an important way of conserving native birds that are under threat from introduced pests like rats, stoats and possums.
Zealandia in Wellington is the world's first fully-fenced urban sanctuary, where native birds inside are doing really well, but some species struggle once they hop the fence and have to fend for themselves.
A team from Victoria University is using a combination of cameras and birdsong recorders at 50 different locations around Wellington to map where species like the Hihi (Stitchbird), Tīeke (Saddleback), and Kākāriki (Red-crowned parakeet) do best.
The idea is that if we know where the birds thrive we can engineer more bird-friendly communities in the future.
Recording the birdsong is relatively easy but trying to identify an individual bird call in over 100,000 hours of recorded audio, which includes other noises like traffic, dogs barking dogs, alarms and doorbells, is no easy task.
And that's where the artificial intelligence comes in.