Optimal Ripeness Classification of the Philippine Guyabano Fruit using Deep Learning

The Philippine Guyabano fruit has a sweet flavor and a soft heart-shaped spine. The existing process of evaluating and rating the quality of the fruit is done manually, using the traditional method, which needs additional effort on the part of graders and o in a visual and tactile assessment. In today's world, the Convolutional Neural Network (CNN) is a well-known used deep learning architecture for vision applications, as well as one of the most widely used classes of image classification and identification models. The researchers investigated various CNN structural designs to select the optimal prediction classifier based on their performance, including VGG19, InceptionV3, and mobilenetV2. The purpose of this study is to use CNN to execute an image classification strategy on data obtained from Philippine Guyabano fruit. The overall efficiency of the model was measured by evaluating the test results using the accuracy and loss function estimates given by the model. In comparison to InceptionV3 and MoboleNetV2, VGG19 provided the greatest model for fruit categorization accuracy among the three models, with a score of 99.25%. As a consequence, the VGG19 model produces the top transfer learning outcome, indicating that it is the ideal fit model among InceptionV3 and MobileNetV2.