Developing a Tool to Classify Different Types of Fruits Using Deep Learning and VGG16

Abstract

In this paper, we present two methods for the classification of fruits of Bangladesh from image processing techniques. We have used deep learning convolutional neural network in our model and VGG16 in another model. From both models, we have found 99% accuracy. Initially, we used only five classes (apple, orange, jackfruit, watermelon, banana) for building these models. Evaluating our model gives us accuracy on the test dataset and by inputting one fruit image our model predicts the fruit what it is. We have checked and experimented with our model several times that it can detect fruit accurately from single fruit images. If our model goes through further improvement, it can be an application that will help shopkeepers or farmers on fixing price calculations on both online and offline platforms.