Classify Tomato Fruit Images Using Convolutional Neural Network (CNN) Method

Abstract

The agricultural and forestry product processing industry is developing very rapidly. The stage in the process of processing plant products is selecting products according to their quality, for example fruit ripeness. The process of selecting agricultural and horticultural products is often largely based on human perception of the color composition of fruit images (manual selection). The weakness of manual fruit classification is greatly influenced by the subjectivity of the classifier. To reduce subjectivity and manual methods, a Convolutional Neural Network (CNN) deep learning method is needed. Tomatoes are chosen for image classification based on color diversity, aiding accurate recognition through distinct ripeness variations. The aim of this research is to detect and recognize tomato images and determine its accuracy value by applying the CNN Deep Learning method. The first stage is preparing the required tomato image data set. The second stage is preprocessing and sorting the tomato image. The third stage is model formation and system training. The last is to carry out system testing. This research uses 14 tomato images which are used as testing data from 56 tomato images used in the training dataset. Testing tomatoes produces an average data testing accuracy value of 97%.