WeedScan 2.0 reaches a major milestone

WeedScan 2.0 has reached a significant milestone, with model training now underway using one of the most comprehensive curated weed image datasets assembled for an Australian identification tool.

Behind this milestone is a substantial national effort to strengthen one of Australia’s most important digital biosecurity tools. WeedScan is a free mobile app that uses artificial intelligence to help land managers, farmers, community groups and the public quickly identify weeds and access reliable management information. By putting expert knowledge in the hands of users, wherever they are, the app supports earlier weed detection, faster response and more informed decision-making.

For WeedScan 2.0, the CISS team has curated close to one million images, all of which have passed independent external quality assurance. The upgraded model will recognise more than 900 plant species, nearly doubling the number available in Version 1.0. Images have been carefully selected across priority weeds and native weed lookalikes, helping to improve identification accuracy in real-world situations.

 

While this work represents a major step forward, users will not see the new functionality improvements until WeedScan 2.0 is released. Alongside the expanded and quality-assured dataset, we have also been improving the user experience and performance of the app itself. For example, WeedScan 2.0 includes a clearer results screen, and improved pathways to move from identification to action by surfacing management guidance and group functionality.

The current focus is now on rigorous model training followed by extensive User Acceptance Testing (UAT). Volunteers across Australia’s jurisdictions will be providing valuable in-kind support through testing and feedback to ensure the app performs reliably across regions, seasons and use cases.

WeedScan plays a critical role in Australia’s biosecurity system. As invasive plants continue to spread, rapid identification and access to trusted information are essential. Tools like WeedScan support coordinated national surveillance and empower communities to be part of the solution.

This progress has been made possible through the support of the Australian Government Department of Agriculture, Fisheries and Forestry, Macpac’s Fund for Good program, which is funding the model training phase, the strong collaboration of state and territory partners, and many dedicated individual volunteers, experts and supporters.

We look forward to sharing further updates as WeedScan 2.0 moves through testing and closer to release.