A team of researchers in Mexico, from the Technological Institute of Orizaba and the Autonomous University of the State of Mexico, has presented in the journal Agriculture a prototype that automatically identifies defects in green coffee beans using artificial intelligence.
The device takes 350-gram samples, photographs small groups of beans with two 16-megapixel cameras under controlled LED lighting, then mechanically flips them 180 degrees and photographs the other side. Analyzing both sides improved defect detection by 25.7% compared with photographing only one side.
The system was trained on 9,600 images covering 11 defect types (black, sour, broken, “elephant” beans and others) plus normal beans. Of the four neural networks tested, YOLOv8 performed best, with 97.4% precision and 99.6% recall. The authors note, however, that one sample takes 38 to 40 minutes, so the prototype is not yet an industrial sorter.
Discover the coffee selection available at Mister Tabaco.
Source: Daily Coffee News