Recognition of Fruit Grading Based on Deep Learning Technique

Abstract

Fruits, with their delightful taste and rich nutrient content, play a crucial role in human health. Their classification is vital in agriculture and influences pricing in supermarkets. Efficient fruit detection is critical for import or export due to physical environmental effect. This project implements the deep learning technique of YOLOv5s and YOLOv8n models to classify fresh and rotten fruits. Besides that, there are two classes as fresh and rotten for grading three fruit varieties involved such as apples, bananas, and oranges.  The 600 sample images are collected from Kaggle and annotated using the Roboflow software. Overall, the result of the proposed project is evaluated using the metric mean Average Precision (mAP). The mAP of YOLOv5 is 98.99% and YOLOv8 is 99.36%. Hence it proves YOLOv8 model performs better is in its identify and classification of the fruit grading.