Automated Grading of Grapes Fruits Based on Internal and External Quality

Abstract

The harvesting time of fruits significantly affects their quality, flavor, and market value. Traditionally, this process relies on subjective human evaluation, which is both timeconsuming and subjective. However, a solution to this challenge is offered through the application of deep learning. This study introduces an approach that utilizes a Customized Convolutional Neural Network (CNN) and the InceptionV3 architecture to differentiate between three stages of grapes. CNNs have proven to be effective tools for automating the assessment of fruit quality by analyzing visual features such as color, shape, and texture. The architecture incorporates multiple convolutional and pooling layers to extract hierarchical features from images, enabling the identification of subtle differences indicative of various quality stages. A dataset named Quality Grading Dataset was created, and the accuracy of various models was assessed: VGG16=96%, VGG19=98%, ResNet50=98%. The accuracy for InceptionV3 is reported to be 98%.