Agrintesa, Sermac and Bioretics: using AI in the agrifood sector
Agrintesa, one of the leading food and vegetable companies in Italy, uses AI to select part of the more than 440,000 tons it produces each year. Established in 2007 to bring together a community of 4,000 agricultural companies that are mainly small- and medium-sized, this Faenza-based cooperative has 200 employees and revenues amounting to more than 300 million euros. With the consulting services of Ser.mac, a specialist in fruit and vegetable processing machinery, and its partner Bioretics, which has developed a visual recognition technology, this company from the Emilia-Romagna region of Italy has now automated every stage of production, which has led to a significant improvement in product quality and production times.
What made you opt for AI?
“Over recent years the market has been evolving very rapidly and traditional technologies, which have been in place for about 20 years, are struggling to keep up with new standards: more juicy fruit, of the right size, that is not damaged. Furthermore, export-bound products must have the right stage of ripeness to avoid their compromising the entire load. These needs made us look for alternative fruit processing methods. With the help of Ser.mac, our long-term supplier, and its partner Bioretics we developed an apparatus that uses AI to scan, recognize and sort kiwifruit and plums very accurately based on their required characteristics. This is taking us towards the goals we set ourselves: more quality, faster. Yet it is precisely because of intelligent visual recognition that the machine learns continuously, improving its capabilities day after day, and in the future this will enable it to maintain a significant competitive advantage”.
How does the technology work?
“In 2017 we added two new machines developed by Ser.mac with the consultancy of Bioretics into our production process. Each fruit is scanned and analyzed using HDiA (High Definition Innovative Agrovision) visual recognition technology that is based on a Google Protocol Buffers data management mechanism. The algorithm, which has been patented by Ser.mac, recognizes weight, diameter, color and quality so as to sort the fruit into one of three production standards (as well as various levels of Premium products). The visual recognition system uses AI to identify the characteristics and outside defects of the fruit thanks to a continuous learning process. With machine learning, the more the tool is used, the more it will become an ‘expert’, improving its own analyzing skills over time”.
What were the results?
The main results were better and more efficient industrial processes. Until 2017, the only step that was not automated was selection, which is the most delicate one. With the introduction of visual recognition using AI, after an initial testing phase, the results were obvious: there were fewer customer complaints about the quality of the product and the process was faster and more efficient. Within two years, once the setup is completed, we expect to also see a consistent improvement of about 10% in the gross operating margin (GOM) for us and our partner producers”.
Would you recommend it to other entrepreneurs?
“After 10 years of working with Ser.mac, we saw AI as being an opportunity to grow, adopting it in a testing phase back in 2017. I believe this is a challenge that all entrepreneurs, big and small, need to embark on if they want to set themselves apart in today’s marketplace with technological innovation. We need to be very aware that it can be a demanding process but, if done at the right time and in the right way, this can offer a company a huge growth potential”.