16 Dec 2020

GOOGLE’S AI TECHNOLOGY TO IDENTIFY ANIMALS IMPACTED BY BUSHFIRES

Artificial intelligence and an army of new sensor cameras will be used to track the recovery of animals impacted by bushfires in one of the most extensive post-fire surveillance programs ever undertaken in Australia.

WWF-Australia and Conservation International, supported with a US$1 million grant from Google’s philanthropic arm Google.org, have launched An Eye on Recovery, a large-scale collaborative camera sensor project.

The project will install more than 600 sensor cameras to monitor wildlife in landscapes impacted by last summer’s bushfires, including the Blue Mountains, East Gippsland, Kangaroo Island, and South East Queensland.

Nearly 3 billion native animals were in the path of the devastating fires and 119 threatened species were identified as needing urgent intervention. The scale of damage is so severe that researchers are still in the field, one year later, conducting ecological assessments.

An Eye on Recovery will use Google AI technology to boost this process, helping researchers to locate surviving wildlife and determine where recovery actions are needed.

The first cameras have been installed on Kangaroo Island – where fires consumed half of the island – to monitor species like the critically endangered Kangaroo Island dunnart.

More than 90% of the dunnart’s habitat was scorched in the fires, but there are signs of hope with the new cameras already capturing an image of the mouse-sized marsupial in Flinders Chase National Park.

Installing sensor cameras on Kangaroo Island for An Eye on Recovery project
© WWF-Australia / Slavica Miskovich
A Kangaroo Island dunnart captured on camera at the new site
© Kangaroo Island Land for Wildlife

“These cameras will allow us to put hundreds of pairs of eyes into bushfire landscapes to locate elusive species like the dunnart. This will give us a better understanding of what animals have survived and where we should target our recovery actions,” said Darren Grover, WWF-Australia’s Head of Healthy Land and Seascapes.

Images captured by the cameras will be analysed by Wildlife Insights, a ground-breaking cloud platform that uses artificial intelligence and machine learning technology developed by Google to identify species.

“Using Google’s AI technology, Wildlife Insights helps biologists automatically identify and share sensor camera images, which reduces time spent manually sorting through images to find that rare dunnart sighting,” said Tanya Birch, Program Manager, Google Earth Outreach. "On average, human experts can label 300 to 1,000 images per hour. With the help of Google AI Platform Predictions, Wildlife Insights can classify the same images up to 3,000 times faster, analyzing 3.6 million photos an hour."

“An Eye on Recovery will help eliminate the guesswork on how and when different wildlife species repopulate areas affected by fire. The sensor cameras will be the eyes capturing images of wildlife coming back to burned areas; Wildlife Insights will be the brain interpreting and analysing this data as it comes in,” said Jorge Ahumada, Executive Director of Wildlife Insights and Senior Wildlife Conservation Scientist at Conservation International.

Wildlife Insights has been trained to identify more than 700 species, but this is the first time the platform will be tested on Australia’s unique wildlife.

To teach the AI to identify Australian species, WWF-Australia is calling on the public for help.

Anyone with sensor camera images of native wildlife, especially in bushfire regions, is encouraged to share their pictures to speed up the teaching process.

Western grey kangaroo caught on sensor camera
Western grey kangaroo caught on sensor camera © Anke Seidlitz / DBCA / Murdoch University / WWF-Aus
Brush tail rock wallaby caught on sensor camera
© NSW National Parks & Wildlife Service

WWF is calling for sensor camera images to teach the AI to identify Aussie wildlife

“Like humans, AI models get better at recognising and identifying animals if they can look at hundreds or thousands of images,” said Mr Grover.

“Whether you’re a researcher with a suite of images sitting in a data storage system or a hobbyist with a sensor camera in your garden to monitor local wildlife, everyone can play a part in this project.”

People with sensor camera images to share can contact WWF-Australia via eyes@wwf.org.au