AI Sorting Machine Detects 7 Potato Defects at 95% Accuracy

A German startup named Carifo has developed an artificial intelligence–driven machine for potato sorting. The system processes up to 10 tonnes of potatoes per hour and achieves a sorting accuracy of 95%. It was engineered specifically for small and family-run farms.

Origin of the Innovation

The concept originated with Benedikt Kessler, who grew up on a potato farm and experienced firsthand the labor-intensive challenges of manual sorting—hours of work conducted amid noise and dust, with each tuber inspected individually. While studying at the Technical University of Munich, Kessler began developing an AI-based system to streamline this process.

Defect Detection Capabilities

The AI model was trained on more than 100,000 images and can identify seven distinct types of defects, including rot, cracks, wireworm damage, and foreign objects. The machine is capable of sorting unwashed potatoes despite variations in appearance caused by soil type, geographic region, and storage conditions. According to Science Mail, the system can rapidly adapt to the specific potato varieties required by each individual farm.

Design Optimized for Small-Scale Operations

The equipment was designed to be more compact, less expensive, and easier to maintain than conventional sorting lines. Its standardized design allows for straightforward repairs and seamless integration into existing production lines. Kessler developed the technology as part of his master’s thesis, collaborating with co-founders Johannes von Witzke and Felix Beck. Further development took place through the UnternehmerTUM incubator, and initial prototypes were tested in a high-tech workshop. The company has now begun entering the market.

Science Mail previously reported that Kazan had developed a more suitable method for potato cultivation.