Google Australia and the Queensland University of Technology (QUT) have partnered to develop an AI model that uses automatic audio detection to identify and monitor bird species. This collaboration is part of Google’s Digital Future Initiative, announced back in November 2021, a five-year investment in Australian infrastructure, research and partnerships.
Australia’s natural ecosystems are under threat due to the rise of invasive species, climate change, bushfires and more – and this AI technology is helping experts make more informed, efficient and accurate decisions about conservation and land management.
The AI model was developed using data from the Australian Acoustics Observatory (A2O), a national network of recorders that have captured more than 17 million hours of raw audio since 2019. Traditionally, experts have manually reviewed recordings to identify bird sounds, which is a painstaking and time-consuming process.
The AI model can automatically separate, enhance and completely isolate the sounds of individual bird species while filtering out surrounding noise such as wind and insects. This makes it much faster and easier to identify bird species, and can also be used to track their movements and populations over time.
The model has been trained on recordings of the Glossy Black Cockatoo, a threatened bird species found along the entire east coast of Australia as well as on Kangaroo Island in South Australia. The model has also been shown to be effective at identifying other bird species, including the Kookaburra, the Australian Magpie and the Yellow-tailed Black Cockatoo.
The AI model is open-sourced, which means that it is available for use by researchers and conservationists around the world. This will help to accelerate the development of new tools for biodiversity monitoring and conservation.
The results so far are impressive and you can see from the audio input ‘field recording’ the subsequent result identifies a range of different sounds. These sounds include Wind, Insects, Mammals, King Parrot, Glossy Black Cockatoo, and The brown hornbill.
Professor Paul Roe, Head of QUT’s School of Computer Science and the Lead Researcher at A20 said Google’s model was producing impressive results, even picking up very faint calls of the bird that would often be easily missed by more traditional methods of identification.
Dr Daniella Teixeira, ecologist and Research Fellow at QUT, said Australia’s natural ecosystems are under threat due to the rise of invasive species, climate change, expanding human development, deforestation, bush fires, and many other factors.
This new partnership between Google and QUT will help researchers to produce high-quality range maps for Australian species and enable conservationists and ecologists to locate species of interest easily.
Together, researchers will look to extend these tools to identify both threatened (such as koalas) and invasive species – such as cane toads, Asian house geckos and Indian myna birds.