Post-harvest sorting of potatoes is a key step in ensuring quality, preventing degradation of stored stocks and reducing subsequent losses. Traditional manual sorting methods are laborious, subjective and inconsistent, especially in large operations. This study presents an approach to potato detection and classification using the YOLOv11 “You only look once” architecture. The application is in a conveyor sorting table environment for post-processing. Two custom datasets were created for clean, washed potatoes and for potatoes in harvest condition (unwashed). The datasets included three classes: good potatoes, bad potatoes and other objects (rocks, dirt, leaves and dirt). Images were collected under a constant light source in the natural movement of the conveyor to reflect real sorting conditions. Models of all sizes from the Yolo11 family were trained to achieve an optimal high average accuracy and model complexity. Inference speed suitable for real-time deployment was also taken into account. The results demonstrate the strong generalization capability of YOLOv11 in complex agricultural environments, enabling automated and objective quality classification for potato sorting lines.
Štursa et al. (Thu,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: