What is automation good for? Harvesting more broccoli than human laborers can, according to Upp, a Shropshire, U.K.-based agtech startup that’s using computer vision AI plus farm-sized proprietary machinery to expand crop yields.
Its pitch is not only that its specialist, AI-driven harvester will make it more efficient to pick a familiar crop but also that the process will reduce waste — by being able to extract more nutritious protein from a field of broccoli without needing an army of extra human workers to do it.
Upp says the smart machinery it’s developing will enable broccoli farmers to harvest more of the plant than they feasibly could using human field laborers because the AI-plus-tractor-tool combo will do it all: Fully automating the spotting, cutting, lifting and carrying, at a rate of up to 3km/h.
This AI-driven approach allows for farmers to “upcycle” the 80% of the broccoli plant (i.e extra stem and leaves) that’s normally left as waste on the field, per Upp, and sell that as a additional product that can be processed into a form it suggests is comparable to pea protein.
The startup’s concept system, which CEO and co-founder, David Whitewood, tells TechCrunch it’s been developing with help from technologists at the University of Lincoln, involves a tractor kitted out with a 3D camera and an on-board computer running a computer vision AI model that’s been trained to identify when broccoli heads are the right size for picking (with better-than-human accuracy, is the claim), along with a proprietary (patent pending) tractor-pulled cutting-and-harvesting tool.
“The job of harvesting broccoli is — firstly — you’ve got to recognize which heads are ready to be harvested. So we’ve been cooperating with the University of Lincoln’s agri products team who’ve been developing the machine learning and AI,” he explains. “We’ve been testing a whole bunch of cameras with them and dealing with the difficult problem of occlusion [where leaves may partly obscure the camera’s view of the broccoli head].
“They’ve using a depth-sensing camera with the 3D piece in it to determine the size of that head. Because we don’t cut every head — we only cut the ones at the right size as demanded by the supermarkets… That then says ‘cut’ and that sends a signal to our on-board computer and then we actuate our patented mechanism that grabs the plant — which would be the same as a human grasping the plant stem — and then a very sharp knife flies in and cuts it in a fraction of a second. And then the plant is lifted away.”
The extra edible plant matter harvested in this way isn’t intended for supermarket shelves — where the stringent cosmetics standards grocery retailers typically apply to their suppliers is a major contributor to food waste by refusing to stock less than perfect looking fruit and veg — rather the idea is for it to be processed into a protein- and nutrient-rich ingredient for selling to the food industry.
Upp envisages the dried broccoli protein being used in a range of products — from sports-style protein drinks to pre-prepared meals and baked goods.
The bits of the broccoli it’s targeting for upcycling are 30% protein by dried weight, per the startup’s website, and also packed with nutrients (vitamin A, B, C, E, K, calcium, iron, potassium, phosphorous, zinc) — as well as being high in fiber.
Upp does not appear to have had any trouble getting early interest from the food industry for the upcycled edible plant-protein — with Whitewood noting it already has a trio of industry partnerships inked (he can’t yet name names but says one is a global “functional drinks” giant; another is a big UK food brand; and the third is a specialist confectionary bakery).
“They’ve very interested in the health aspects of broccoli,” he goes on. “They’re interested in the fact that it’s clean and sustainable… So they’re excited, shall we say. I don’t think we’ve got a problem with a market for it — once we’ve got it off the field.”
On the processing piece, Upp is working with experts at the James Hutton Institute in Dundee to figure out how best “to recover the fractions from that plant that makes it suitable for the food industry primarily”, per Whitewood.
Zooming out, Upp is developing what it bills as a specialist “circular plant protein” business against a backdrop of growing demand for alternative, plant-based proteins as the food industry looks for ways to shrink its reliance on animal-derived proteins in order to reduce its carbon footprint — with global pressure on farmers and food companies to hit climate targets.
Hence the startup is projecting that its AI-harvested broccoli protein could grow into a multi billion-dollar market in the coming years.
On the marketing side it claims an added environmental upside — suggesting broccoli protein is cleaner than pea protein (being 4x less carbon intensive to produce), while also arguing it avoids the deforestation problem that’s tainted the reputation of soya crops. So the pitch is this is an even greener plant protein.
One potential PR wrinkle is there will inevitably be some (human) worker displacement as a result of automating the harvesting of broccoli.
Whitewood says the system replaces about seven field workers — but he notes that “warm bodies” are still needed in the pack house to package the broccoli products for retail. “Seven hard to get people,” he adds, sketching a picture of the gruelling work field laborers typically have to do and arguing these aren’t the kind of jobs anyone is going to miss. “Nobody wants to do this work. Even in China and India they’re struggling to get people to do this… It’s the 21st century and we’re still expecting people to do this. It’s just crazy.”
While the 2022-founded startup’s tech has been developed to the concept stage it’s gearing up for the next stage — to hone a robust technology that can be commercially deployed — with a series of “field-to-protein” pilots planned this year in the U.K., Spain and California.
It’s expecting to start commercial production (and generate its first revenues) in late 2024 — projecting revenues will exceed £50 million in its three pilot markets in 2027.
The business was established last year as a spinout from another U.K. agtech business called Earth Rover — where Whitewood had been CEO before moving over to Upp as a co-founder when they decided to separate into two distinct businesses.
Today the startup is announcing a £500,000 pre-seed investment from Elbow Beach Capital, a decarbonisation, sustainability and social impact investor, to fund the field trials — ahead of planned commercial deployment later next year.
Whitewood says the first commercial use of the tech will likely be in Spain or the UK, owing to seasonality, before Upp moves on to pitching California’s broccoli growers on automated crop yield optimization.
Why hasn’t anyone done thought about extracting more of the good stuff from broccoli plants before? Whitewood says people have been thinking about the potential to do this for over a decade but he suggest it’s just “really hard” — given the selective harvesting required, as well as the need to separate out the harvested crop, with part (the broccoli crown) going to supermarkets (to be sold fresh) and the rest requiring additional processing.
“It sounds simple — a lot of people have tried and a lot of people have failed,” he suggests. “It’s only since you’ve got a specialist harvester that can handle all the bulk that suddenly you can start to deal with the rest of it. You need automation — and it needs big automation. Little robots aren’t going to deal with crops of this scale, this bulk… You need farm-sized machinery.”