Revolutionary technique identifies individual voices
Artificial intelligence that identifies individual koalas by their vocals promises to revolutionise our understanding of the iconic – and now endangered – Australian marsupial.
Koalas were placed on the endangered list this year following a catastrophic drop in numbers due to a perfect storm of habitat destruction, diseases, bushfires and floods.
Wize Dynamics was one of four businesses who received a NSW Government grant of $100,000 to pilot the use of new technologies in detecting and quantifying koala populations for the benefit of preservation.
Wize Dynamics was formed by University of Wollongong graduates Bilal Arshad and Elliott Pilton. They teamed up with UOW researchers Hoa Dam, Kimberly Maute, and Katarina Mikac to undertake a study into the development and suitability of artificial intelligence to identify individual koalas from audio recordings.
Chief Operations Officer Elliott Pilton said Phase 1 was now successfully completed.
“We succeeded in being able to reliably identify an individual koala with 94 percent accuracy, under the leadership of Bilal Arshad, Director at Wize Dynamics,” he said.
“Our solution records koala bellows without disturbing koalas, and other species.
“The design is inspired by the fact that very few physical monitoring solutions for nocturnal and arboreal species are truly passive in nature and cause no concern to the welfare of the animal, as well as being inexpensive and easy to use.”
He said that about 20 studies along the Australian East Coast were recording koala bellows, creating about 100,000 hours of audio in a single mating season, from October to December.
“Since each hour takes about two human hours to log – and that’s before any analysis – the existing methods are time-consuming and expensive,” he said.
“In addition, it can take up to two years between the collection of data and government action. We simply don’t have time on our side.”
Wize Dynamics is now applying for Phase 2 of the NSW Small Business Innovation and Research Program, which will allow tracking of individual animals through their bellows using multiple recorders.
Grants are worth up to $1 million with expressions of interest due by Friday, 8 April.
About Wize Dynamics
Wize Dynamics started in 2017 after a collaboration of UOW graduates – Bilal Arshad, Elliott Pilton and Kaiyi Cao – and an incubation in the University’s business incubator iAccelerate.
The business has developed a trap camera named ‘WIICD’ (Wildlife Intelligently Identified and Counting Device) that uses digital machine learning to identify, transmit and track animals, from visual data(images and video).
This allows calculation of animal movements and population density and has already been used to study feral deer in the Illawarra.
