AN artificial intelligence (AI) grant from global tech giant, Microsoft, has given a massive leg up to homegrown Australian ag tech startup company, InFarm.
InFarm is the only private company among six Australian entities to receive grant support through Microsoft’s AI for Earth program.
The small company, which was established in 2017 at Goondiwindi in southern Queensland, specialises in drone imagery, weed identification and precision agriculture services for the rural sector.
InFarm managing director, Jerome Leray, said the Microsoft grant would be used to ‘train’ the business’s AI systems, focusing on detecting two problem weeds – Harricia cactus and Feathertop Rhodes grass – that mainly affected south west Queensland and north west New South Wales.
Feathertop Rhodes grass has developed high levels of resistance to agricultural chemicals and is a major problem for cropping farmers.
“With Feathertop Rhodes grass, what we are doing is creating a system that uses aerial images from drones to detect where the grasses are. Then we provide a map that farmers can upload into their tractors to do targeted spot spraying,” Mr Leray said.
Harricia cactus is mainly a problem in grazing situations. It is extremely hard to kill and has a long seedbank life.
“With Harricia cactus, we are using the AI system to detect those weeds. Once we have that we can do a count. Also, because they take so long to manage we can see and evaluate how different management systems are going,” he said.
“That is what a lot of the NRM (Natural Resource Management) groups, Landcare and government agencies are looking to deal with because it is such a problem weed. But they don’t have a good idea of where it is, how much there is and how the different management strategies are working. With the ability for us to map and identify where they are, we can provide that information to these organisations and farmers.”
Mr Leray said InFarm company had until November to use the grant which would vastly speed up the development of its AI systems.
“Essentially what they are doing is giving us access to high performance computers which allows us accelerate our training and develop and test the system. Instead of taking 12 to 18 months it will take us three to four months in terms of iteration,” he said.
“There are times when we are training the AI system where it can take us two weeks, but using their computers we will be able to do it in two to three hours. It is something we don’t have access to in regional Australia.
The other Australian recipients of Microsoft AI for Earth grants were Monash University, Griffith University, Queensland University of Technology, the Australian Wildlife Conservancy and Bush Heritage Australia.