Scientists protecting Kakadu National Park are calling on citizen scientists to take part in a digital fishing challenge to help generate the largest dataset of fish species in the world.
In partnership with Microsoft, the Australian Government’s Supervising Scientist Branch has developed an AI-based solution to track underwater species, which will help researchers and Indigenous rangers safely monitor billabongs without fear of croc attacks.
Minister for the Environment Sussan Ley said scientists looking after river health need the public’s help to label more than 500,000 images of fish.
“This dataset will make it far easier to train artificial intelligence to identify different fish species, creating a tool which will support the protection of the dual-world heritage listed Kakadu National Park,” Minister Ley said.
“We are asking members of the public to simply log on and annotate the images. Some of the species you may encounter include the barramundi, the chequered rainbowfish and the penny fish.
“This is a great way to have a positive environmental and scientific impact from your sofa.”
Supervising Scientist Keith Tayler said with the help of the public it is hoped the tool will eventually be used by Indigenous ranger groups to safely undertake long-term environmental monitoring at Kakadu.
“Fish monitoring was previously tackled by teams of up to 15 people working in the field for weeks at a time, who had to use nets to sample fish or count them through see-through panes in a bubble boat,” Mr Tayler said.
“However increased crocodile activity in the area has made that a perilous proposition.
“This technology will mean ecologists and rangers no longer need to manually process videos of fish which frees up time for other research to help protect and rehabilitate ecosystems.”
Microsoft Chief Technology Officer Lee Hickin said using cloud computing and artificial intelligence to automate fish stocks would help scientists around the globe.
“Our collaboration builds on a similar AI-based solution developed in 2018 to identify fish in Darwin Harbour that was the result of a partnership between Microsoft and the NT Department of Primary Industry and Resources fisheries team.
“The framework developed for Kakadu can be re-deployed anywhere, from Antarctica to the Great Barrier Reef. The dataset, called BRUVNet, will be made publicly available to fisheries scientists around the world.
“By using AI and Deep Learning to automate the data processing and analyses in the back end, we will have a highly complex, easily deployable and scientifically rigorous environmental monitoring tool.”
The Supervising Scientist Branch within the Department of Agriculture, Water and the Environment carries out environmental research and monitoring to protect Kakadu National Park from the effects of past uranium mining.
To take on the digital fishing challenge visit https://www.bruvnet.org