An Automatic Fruit Image Classification System

Abstract

Classifying fruits and vegetables is still challenging in daily production.  Deep Convolutional Neural Networks (DCNNs) have made significant progress in solving prediction problems, such as object recognition, scene interpretation, and semantic segmentation, frequently outperforming humans in accuracy. In this study, we provide an effective fruit classification system in digital images utilizing deep learning techniques. By training the system on images of three different fruit categories: grape, citrus, and pomegranate, a deep learning strategy based on convolutional neural networks (CNNs) has been constructed to classify the item (fruit). We created an algorithm that automatically extracts and uses features from images in training.

The dataset used is 600, for training 80% were used, while the remaining images were used for testing. Based on our experiment, we discovered that 60x60 pixels is the ideal input image size, and 100 epochs is the perfect number. The accuracy of the test photos reached 97%, and the results are excellent. The findings demonstrate that the suggested methodology improves fruit classification ability overall.