A startup with ties to the Technical University of Munich (TUM) has built an artificial intelligence (AI)-powered optical recognition system capable of sorting up to 10 tons of potatoes per hour with a 95% rate of accuracy.
Manual potato sorting is labor-intensive, with workers spending hours inspecting individual potatoes in noisy, dusty conditions, which is what inspired the team from the startup Karevo to develop its AI-powered optical recognition system.
Source: Karevo
To develop the system, the team trained a model using more than 100,000 images, thereby enabling the system to identify damaged potatoes with a high rate of accuracy. According to its developers, the machine can successfully detect foreign objects as well as defects like rot, cracks and wireworms. The system also works on unwashed potatoes.
Further, Karevo’s AI-based sorting system is capable of handling the variations in unwashed potatoes incurred by regional, soil and storage conditions. Likewise, according to company the machines are smaller and more affordable than current potato sorters, with a modular design that simplifies maintenance and integration into existing systems.
