A Deep Learning Approach to Identify Fresh and Stale Fruits and Vegetables with YOLO
M. Sai Sree Akshitha Reddy, N Aishwarya
Abstract
Consuming fresh fruits and vegetables is very important not just because they taste good but are also a rich source of vitamins, minerals and dietary fiber. About 20% of the fruits and vegetables that are cultivated for human needs are wasted due to spoilage. Hence, it is important for a farmer to remove the damaged ones so that the other ones do not get spoiled. Identification of fresh fruits and vegetables also helps a person in picking good ones from the shop. This project focuses on identifying if a fruit or vegetable is fresh or stale using two different versions of YOLO namely Yolov4 and Yolov5. Images of six different types of fruits and vegetables: apple, banana, orange, tomato, capsicum and bitter gourd are considered in this work. As there are 6 varieties involved, it is a multiclass classification model with 12 classes. The overall results show that the trained Neural Network achieved a classification accuracy of 95.9% for Yolov4 and 99.9% for Yolov5 on a dataset of 1200 images.