The Queensland Department of Primary Industries (DPI) has teamed up with commercial partners to create artificial intelligence designed to provide more information about calving rates and disease infestations on large-scale properties.
As the principal Agtech scientist with the DPI, Paul Stewart works with the industry to understand the needs of graziers across northern Australia.
“Where we saw the need is … to try and get ahead of these emergency animal diseases that are coming up,” he said.
“[Also] you may only muster [cattle] a couple of times a year, and it’s important to have eyes on your herd all the time and be able to see when cows are calving.”
Cameras are being used to monitor cattle as part of a DPI trial with two commercial partners. (ABC Landline: Cameron Lang)
As part of the trial, trailers equipped with motion-detection cameras have been set up around cattle watering holes.
One early obstacle has been getting the machines to recognise the animals and diseases they are meant to detect.
InFarm, a firm that specialises in AI and advanced technology for agriculture, have been working with the DPI to differentiate between animal types.
“We took footage of the cows, and then we did a thing we call ‘Tinder for cows’. Is this a cow, or is this a calf? We swipe left and right,”
managing director Jerome Leray said.
At the same time, Think.Digital CEO Kat Bidstrup and her company have been using a machine learning algorithm to spot diseases.
But it’s been a difficult process, as Ms Bidstrup has had limited access to cattle with the relevant diseases in Australia.
Jerome Leray has been working with the DPI on the project. (ABC Landline: Cameron Lang)
“We put out a worldwide call …[and] had some great support from the EU [about] foot-and-mouth disease,” she said.
The Australian government have also been helpful in providing images.
“But we actually ended up having to fund a vet in Indonesia to go out into the hot zone and take some photos to build up our database,” Ms Bidstrup said.
Calves … or pigs?
Mr Leray has encountered a different set of issues in his mission to detect calves.
“We started to get this weird clustering of calves [in the footage] that just didn’t look like anything else — and we actually determined that it was pigs” he said.
“So that was probably something that was a little bit unexpected.
One of the biggest challenges for the project is connectivity, which can be poor on large properties. (ABC Landline: Cameron Lang)
“Calves that are newly born, calves that are a lot older, with hills in the background, without hills in the background [were also a problem for the AI].
“If an AI has never seen certain features, it won’t be able to generalise.”
AI scanning for sickness
Australia is currently free of foot-and-mouth disease and lumpy skin disease, but the discovery of either would trigger a significant response here.
Information is crucial, and the situation could become complex if a disease like lumpy skin is discovered on a property in Australia for the first time.
Once monitoring equipment is set up on a wider scale, questions may arise about data ownership, data access, and the process of farmers themselves discovering new diseases on their properties.
But these are all issues the DPI is looking to address.
The technology is still in its testing phase, with disease detection relying on images. (ABC Landline: Cameron Lang)
“[What] we’re trying to do is get the jump on any kind of incursion that we may have,” Mr Stewart said.
“The earlier we can get on to it, the more chance we have of controlling and eradicating.”
He stressed the need for a rapid response to disease outbreaks.
“If they can act early on, that’s better — whether it be an endemic disease that they’ve already got, or maybe in one of the emergency animal diseases [like lumpy skin disease],” he said.
Initial trials with future promise
While the process of testing and teaching the AI programs what to detect is ongoing, there are also other issues to overcome.
“Our connectivity hasn’t been great in Australia, so we’re starting to deal with that,” Mr Stewart said.
Kat Bidstrup and Paul Stewart are optimistic about the technology’s future. (ABC Landline: Cameron Lang)
“There’s the problems of having infrastructure, power, and things to actually put cameras on, and that just makes it a bit more of a challenge.”
At this stage, the teams are focusing on mounting cameras around watering holes to ensure the cattle come to them, but the scope is growing, and Mr Stewart is excited about the potential of the technology.
“In the future, we might be able to take it to things like weight estimation and lameness and those kind of things — that’d be pretty exciting,” he said.
Watch ABC TV’s Landline at 12:30pm on Sunday or on ABC iview.
