Job offersmore »

Tweeting Growers

Top 5 - yesterday

Top 5 - last week

Top 5 - last month

Exchange ratesmore »

SunSelect: Cutting yield prediction error rates by half with AI algorithm

Predicting the yield of your crop based on a self-learning algorithm, custom-made for your greenhouse enterprise and driven on artificial intelligence. It might sound like a futuristic scenario, but it's here. Over the last year, US tomato grower SunSelect (150 acres) worked with a beta of Motorleafs’ Yield Prediction System and cut yield prediction error rates by half. The system is now commercially available. 

Predicting yield
SunSelect’s co-owner Victor Krahn explains: “For us (and any commercial size operation) knowing/predicting exactly what quantity of product we'll have each week is one of the top metrics we obsessively track. This data feeds into all critical areas of our operation, from labour management to price negotiation, to buyer relationships and even trickling through to in-store marketing budgets. Like most commercial greenhouses, our ability to predict something so dynamic as yield is almost impossible to get right. Our error rates bounce between 5% to over 35% in some weeks.” 

Over the last eight months, the company worked with a beta version of the Yield Prediction service for commercial greenhouses, developed by AgTech company Motorleaf. It was launched in, their virtual agronomist platform for greenhouse and indoor growers worldwide. “We didn’t think we could still be surprised, but the numbers proved to be more impressive than we expected”, Motorleaf-CEO Ally Monk says. “We’ve seen the number of errors being reduced by 45.9 per cent. For us, that’s the number to watch.”

How does it work?
Time to spill some secrets. How does the Motorleaf prediction work? The system is built on information from the greenhouse and the crop, like location, variety, density, and environmental data. This is combined with visual data, contained by several cameras in the greenhouse and Motorleaf's hardware that collects additional data not normally collected by greenhouses such as light spectrum." With this information, Motorleaf creates a unique algorithm for the greenhouse. “But that’s not the biggest gain”, Ally adds. “In a certain way, a yield prediction is like a soup – everybody uses their own ingredients and creates their own version. The unique part of our system, making it the first in the world as far as we know, is that the algorithm we created is bettering itself – it’s a self-learning yield prediction machine. Based on the algorithm, the data and the yield, the prediction generated by the system gets more accurate every time. That makes it a unique system in the greenhouse growing industry.”  

Victor can add some numbers from the SunSelect experience to make this explanation more concrete. “The latest algorithm is close to cutting our yield prediction error rates by half. It goes without saying it's a game changer that no one saw coming. Motorleaf is now our new standard for predicting yield.” 

Commercial deployment
With the beta being fully tested over the last eight months, Yield Prediction is ready for commercial deployment now. “We can create a unique algorithm for a greenhouse in a matter of weeks instead of months, which was the situation before. Our suite of hardware and software can leverage the data it collects via its own hardware, but can also leverage data that's already been collected by greenhouse clients.”

Mr Krahn continues: “Working with our team of growers and VP of Operations we were impressed at the level of accuracy Motorleaf’s team provided through both touching the plants in the greenhouse and reaching through the cloud for the data our state of the art greenhouses provides every day. The integration was easy, non-invasive and I think shows amazing promise to become the new industry standard.”

For now, the algorithm is available for tomato crops. “But stay tuned; we’re definitely interested in conversations with pepper growers.”
Yield Prediction is not the last we’ll be hearing from Motorleaf. For the future, the company is working on expanding “Currently in most cases valuable data is being collected, but is not being leveraged to its potential by the greenhouse”, Scott explains. He has been working on artificial intelligence in greenhouse crops for Motorleaf and sees a big future for the self-learning systems. “So much in the greenhouse operation depends on the yield and by this system, we’ve opened doors further. Crucial in this is not to get rid of traditional farming, but to learn from the knowledge gained by the farmers and to combine this with the measured data and the learning capacity from the system. All this is done in a secure way of course, to make sure a grower's data pack is not shared with other growers.” 

A next step for the company will be the focus on the limiting points in the growing system. “By combining knowledge and numbers with machine learning, we can hand out shortcuts for the best growing advice. It will take growing to the next level – and add dollars to our clients' bottom line.” 

For more information:
Alastair Monk 

Publication date: 1/18/2018
Author: Arlette Sijmonsma



Other news in this sector:

8/17/2018 Russia: More than 180,000 tons of vegetables harvested in Krasnodar Territory
8/17/2018 US (NV): Private school greenhouse grows for the community
8/15/2018 Can artificial intelligence do as well as farmers in the future?
8/15/2018 Hyper-local cultivation: Restaurant integrates aquaponics greenhouse
8/15/2018 US (WI): 20 hectare greenhouse operation to begin next spring
8/14/2018 French shelter plans to build rooftop greenhouse
8/14/2018 Nanotech company to develop greenhouse film for space missions
8/14/2018 Arctic agriculture: Smart greenhouses and light on demand
8/14/2018 US (HI): 7 hectare tomato greenhouse up for sale
8/14/2018 Hoogstraten has the biggest greenhouse area in Belgium
8/13/2018 “Is the Turkish lira past saving?”
8/13/2018 Cost benefits of greenhouse controllers
8/9/2018 "Tackling difficult issues from a different angle"
8/9/2018 US (AR): Advanced greenhouse will amp up rice research in Stuttgart
8/8/2018 Kadyrov claims Chinese greenhouse vegetables absence of harmful chemicals
8/8/2018 Russian scientists grow wild strawberries in new smart greenhouse
8/8/2018 Greenhouse production in Turkey overview
8/8/2018 US (PA): New “parklet” greenhouse provides green oasis & fresh veggies
8/8/2018 US (MI): Greenhouse gives Indian Reservation access to fresh produce
8/7/2018 Students competing in the Greenhouse Challenge are in the final sprint