Automated Fruit Sorting System using Computer Vision for Retail & Industry

In this paper, we discuss the implementation of an automated fruit sorting system utilizing You Only Look Once-v7 (YOLO v7) for object detection in conjunction with Raspberry Pi for embedded realizations. The system identifies and sorts apples, bananas, and oranges into fresh and rotten categories, thereafter sorting the fruits into corresponding bins with servo-assisted mechanisms. Using Roboflow for dataset labeling, the system attained an average classification accuracy of 85% after training with 950 images. The fruits were continually conveyed with a 12V Johnson Geared Motor powering the sorting mechanism. Planned improvements include the use of spectral and olfactory sensors for multi-fruit detection, as well as increased spatial resolution.