Detecting priority weed species at scale with drone AI
Project summary
Invasive weeds can have severe consequences for biodiversity and threatened species. Finding and treating weeds is costly and time-consuming when they have to be located and identified on foot. The Australian Government is supporting an AI project that provides a quicker, cheaper and more accurate solution to detecting and managing weed species.
The Saving Native Species Program’s Threat Innovations grants were designed to encourage continued investment in longer-term actions to reduce pervasive threats at the landscape scale. In 2024, Climate Change, Energy, the Environment and Water awarded a grant under this program to the Centre for Invasive Species Solutions to pursue the WeedRemeed™ project.
The project builds on existing WeedRemeed™ technology that uses a colour-picking mechanism to detect weeds. The upgrade includes AI-driven image analysis to improve accuracy and include a wider range of weed species. It will enable environmental managers to use drones for large-scale weed assessments, detecting them earlier and targeting on-ground actions like removing and treating them more easily.
Mallee fowl nest photographed by drone at the Bullock Bridge WeedRemeed trial site in South Australia, where Coastal Wattle was targeted for detection during the project trial.
Image: Chris Warrior, Wiru Drone Solutions
How AI contributes to the project
WeedRemeed™ uses AI drone technology to detect weed species at landscape-scale. Drones equipped with cameras distinguish between images of native plants and weed species.
High-quality data is needed to train the AI model to detect weeds and native plant species. This data is being collected using drone imagery across 5 sites in Victoria, South Australia, Queensland and the Australian Capital Territory. These sites were selected because they are representative of a wide range of ecosystems.
Experts in plant biodiversity and weed assessments check the images to determine the species. Some of these verified images are used to train the model. It is then tested against other verified images to determine its accuracy at detecting weeds. The model is being trained to detect weed species including Wheel Cactus, Parkinsonia, and Coastal Wattle.
Project outcomes and next steps
As part of the grant, WeedRemeed™ is being scaled up to be used nationally. The technology will be used by environmental managers and Nature Repair Market users across Australia. It will enable managers to survey large areas more quickly and accurately and target the best areas for weed management. This means limited resources will have the greatest impact.
Surveys of approximately 400 hectares a day are completed with this technology, a significant improvement over traditional on-ground survey methods. This is particularly important in large conservation estates that are remote or hard to access.
As project implementation continues, more quantified on-ground results will become available. As the technology advances, more weed species will be included in the model, expanding its uses.
Find out more
Centre for invasive Species Solutions (2025) WeedRemeed™: Drone AI detection of priority weed species at scale, CISS website, accessed 29 July 2025.
Department of Climate Change, Energy, the Environment and Water (2023) The Threatened Species Action Plan, DCCEEW website, accessed 7 August 2025.
Department of Climate Change, Energy, the Environment and Water (2025) Nature Repair Market, DCCEEW website, accessed 29 July 2025.