AI for biosecurity

Overall Objective

The aim of the project is to develop a AI-based mobile tool that would allow grain farmers to identify a wide range of potentially harmful insects using image recognition technology.

Project Synopsis

Dex successfully used computer vision, a branch of machine learning, to create automated programs that can detect threat insects in photographs.

The computer vision models target most of the insects in Australia’s top 40 pest threats, plus additional insects in the PGAI list of grain pests and the WA government’s declared pest lists. Of 54 significant and high-risk genuses, 43 have a recall accuracy of more than 95%, and 27 have a recall accuracy of more than 98%.

We will operate six connected computer vision models: One is designed to identify which taxonomic order an insect is in (such as Lepidoptera, butterflies/moths), and then pass the image on to one of five more specialised models for orders with significant pest insects. The five specialised models are focussed on Hemiptera, Hymenoptera, Lepidoptera, Coleoptera and Diptera.

We expect our web portal to be operating by November 17th and our app to be admitted to the app store by December.

The cloud-hosted models, the portal, and the app will decrease the time required to identify potential pests, and decrease decision to eradication times. It will put the brain of the entomologist in the hands of the farmer.