A number of exciting concepts emerged from Agri-Tech East’s first hackathon, which was designed to focus the diverse talents of multi-disciplinary teams on intractable problems in agriculture. The overall winner was ‘WeedSpot’, which applies gaming technologies to artificial intelligence weed identification.
Hosted in partnership with Allia Serious Impact, and kindly sponsored by BASF, Barclays and SmithsonHill, the >sudo : grow hackathon attracted participants spanning five decades in age, with representatives from four continents and from many disciplines. Farmers, plant scientists, coders, machine learning experts and software engineers all worked together over the weekend.
The nine teams presented a number of novel solutions to three challenges articulated by the farming community.
Challenge One – Supporting and enhancing traditional approaches to weed control
This topic was introduced by Louis Wells of BASF, who described the challenge of weed control, and in particular Black-grass which infests over 1 million ha of UK arable crops. Over the years herbicide options have become more limited with weed resistance increasing in the population. This creates an opportunity for new thinking and innovation to enhance traditional approaches; from sophisticated weed spotting, to electric weeding, or something else. The challenge is to develop new techniques to enhance what we do today..
Proposed solutions: Two teams took up this challenge and the winner was WeedSpot, which used a gaming engine to create 3D visualisations of weeds. Synthetic visual data (was created to help train the algorithms for weed classification supplemented by real-world data to validate. This deep learning approach also has applications for disease detection by simulating IR or LIDAR data, for example.
Another approach was WeedBeGone which electrocuted weeds using a tractor-mounted ‘zapper’. As well as WeedSpot being crowned overall winner, there was a commendation prize from the judges for Lavender Hill Mob, awarded by Eric Ober of NIAB.
Challenge Two – Making food production more accessible
Vertical growing systems that use water and soluble nutrients instead of soil offer the potential to grow more within a restricted space. Opening up food production in unexpected places. But the system needs sensitive control to ensure optimum growing conditions for different crops in different farming rigs. How could this be achieved?
Proposed solutions: Three teams took on this challenge and the winner was team Grow’s solution for an AI controlled system with an easy to understand dashboard; for example it could use temperature sensors to generate a heat map. It created a connected IT solution for collaborative working across different modules of vertical-farming units, allowing feedback of information about crop development in real-time and transparency with potential customers in the value-chain.
Lavender Hill Mob built a simple dashboard for vertical farming that used a traffic light system for easy identification of the priorities. It also offered potential to look at trends and compare the outcomes to the growing conditions.
Another team, Third Eye, came up with IoT-enabled sensor devices for remote sensing and management of the Aponic platform using technology inspired by gaming.
Challenge Three – Data integration to provide more accurate assessment of growing conditions
Sachin Shende of KisanHub introduced this topic, explaining how farmers capture huge amounts of data from different sources when monitoring and managing their farms. From weather data and crop yield information to soil moisture and drone imagery, there is an increasing amount of ‘big data’ available in multiple formats. However the challenge is how to make the data specific enough to a particular location. For example if the weather stations are 6 miles apart and one records rainfall and the other doesn’t how do decide on the conditions for a particular field?
Proposed solutions: This challenge attracted four teams and the winner was Durian Pi. It combined existing data to identify new correlations and used interpolation techniques on these to predict weather data in-between the 26 weather stations to which KisanHub gave them access. Interpolation often leads to inaccuracies in real-time situations and to adjust for this the team correlated different parameters to help farmers get better predictive data – transforming a basic weather model into agriculturally valuable data.
Other teams included Duso, which developed a wind speed prediction tool to help farmers manage their spraying schedules, and the Water Pirates, who presented an irrigation scheduling platform.
The prize for the best tweet went to Emma Fletcher from Smithson Hill (see below or click here):
— SmithsonHill (@SmithsonHill) April 7, 2018
Comments from the event:
“This is long-overdue. It could be the beginning of a whole new era for putting the technology into agri-tech.” – Jason Hawkins-Row, Aponic
“A phenomenal weekend beyond our expectations! The breadth of enthusiasm and experience of the attendees shone through into some really exciting presentations.” – Paul Hughes, Future Business Centre
“These were the best presentations I have seen in a first run-through at a hackathon.” – Aaron Croucher, PA Consulting, about the first draft pitches on Sunday morning
The judges were: Louis Wells of BASF, Jason Hawkins-Row of Aponic, Sachin Shende of KisanHub and Paul Hughes of Future Business Centre.
The experts and advisors included: Robert Allen of Greenvale AP, Aaron Croucher and Will Wykeham of PA Consulting, Darren Gedge of G’s Growers, Ben Miles of BASF, Stephen Temple of SJ Temple and Sons, and Eric Ober of NIAB.