Cropify takes the guesswork out of pulse classification via computer vision

Unlike cereal and canola crops, which can be relatively efficiently sampled, the process is more complex for pulses and requires the sampler to make a number of subjective calls in regards to seed colour and shape.

However, a new agtech business has come up with a solution that will provide consistent sampling results at good speed.

Anna Falkiner, chief executive of Cropify, said her company was about to undertake a series of commercial-scale pilot projects with leading pulse receival facilities throughout western Victoria and South Australia using the company's high tech product to grade lentils.

"We use a high tech camera system to take an image of the sample, from there we've taken artificial intelligence and machine learning and trained our program to recognise discoloured and defective grain, along with foreign material such as weed seeds," Ms Falkiner said.

"The program's training has been done by highly skilled, laboratory grade assessors and the accuracy is currently at around 98 per cent."

"We've aimed to benchmark ourselves against skilled laboratory classifiers and we're doing well."

Industry reaction has been good.

Brock Tait, general manager at Wimpak, a pulse processor at Minyip in Victoria's Wimmera said the technology would increase grower confidence in sampling process.

"Artificial Intelligence in the grain sampling Industry will level the playing field and eliminate discrepancies," Mr Tait said.

"Grain sampling can be quite subjective from person to person, as well as different delivery sites, this technology should increase grower confidence that when delivering grain, it is correctly and fairly graded," he said.

"We are excited at how rapidly Cropify is evolving and the potential to streamline testing in the industry."

Ms Falkiner said she hoped this would be the case and that the grading equipment would greatly ease the pressure on the pulse receival network.

"You've often got young people on the sample stands that have undertaken fairly quick training, they're under pressure to keep up with a lot of trucks coming through, mistakes will happen, but with this product we can be highly repeatable and objective with what we are doing."

Ms Falkiner said at first the focus would be on lentils, before rolling out across all pulse commodities.

Further down the track she said the company would investigate a product in the cereals space, but said it was a markedly different process than for pulses.

"We thought pulses, with clear visual requirements from the market, would be a really good place to start."

"There is a precedent in the horticulture sector and we thought it would translate well in grains."

There have been challenges, with some weed seeds particularly hard to identify, however with a painstaking period of training the program Ms Falkiner said the Cropify system now had excellent identification of weeds within a sample.

She said the feedback from the receival businesses and packers Cropify had worked with had been positive.

"It is filling in a clear need in the market, not only making it easier on the frontline in terms of shortening sampling stand time and by extension lowering turnaround times."

"At the other end, we're seeing a more consistent product delivered to the customers."

She said it was an exciting period for Cropify, which featured in the evokeAG 2025 startup program.

"It's been busy, but exciting, we're working together with manufacturers here in South Australia, where we're based, to get our all equipment up and running."