Determining the Ripeness of Edible Fruits using YOLO and the OVA Heuristic Model
Analyn N. Yumang, Diego C. Rubia, Kyle Phillipe G. Yu
Abstract
The freshness of fruit can be determined in various ways. And some do it by smelling, some by using touch, others visually, while others even predict the freshness through sound. Among these, the most common and accurate way of determining the freshness of the fruit is by looking at it. Therefore, the authors of this paper made a system with the general objective of automating the prediction of the freshness of two types of fruits which are the Cavendish banana and Carabao mango, by using the You Only Look Once (YOLO) algorithm for object detection and a Support Vector Machine (SVM) classifier that will classify whether the fruit subject is unripe, ripe, or spoiled. Given that the classification problem presented here is multiclass, the One Versus All (OVA) heuristic model is needed to extend the SVM classifier. The YOLO classifier was trained with a dataset that contains more than 100 images, and it was able to detect fruit samples that were invalid and valid. Furthermore, with 33 test fruit subjects, the system’s overall accuracy regarding its classification of freshness to the fruit subjects is 90.9%.