Automated Tomato Sorting and Counting Using YOLOv11 for Industrial and Precision Agriculture Applications

In response to increasingly stringent quality standards, the food industry is progressively replacing manual processes with automation. This study presents a machine learning approach based on YOLOv11 to address the challenge of tomato classification and counting in the industrial sorting process. The model was trained on a custom dataset comprising 1,500 images and over 14,000 labeled instances, enabling it to distinguish among four classes: ripe tomatoes, unripe tomatoes, soil clods or rocks, and plant branches. It was applied to video footage of a factory conveyor belt, where a counting mechanism was implemented to detect and track objects as they crossed virtual lines. This approach facilitates integration with technologies for the automatic separation of fruits and debris in the field, as well as quality estimation by scanning tomato loads, thereby helping to prioritize higher – quality batches. Evaluation over 500 epochs demonstrated strong performance, with high precision and recall across all classes, achieving a best mAP [Formula: see text] value of 84.8% on the validation set. These findings highlight the robustness of YOLOv11 in distinguishing fruits from unwanted elements and its ability to reduce misclassification errors common in traditional sorting methods. This makes the application suitable for both industrial and precision agriculture.