Applying Convolutional Neural Network Algorithm for Classify a Tomato Species

Tomatoes are one of the economic crops that make income into the country. The price of a tomato depends on its shape. However, the tomato species classification currently consumes the workforce, which lows classification rate and accuracy. This research aims to overcome that addressed limitation by applying the Convolutional Neural Network (CNN) algorithm for classifying a tomato species. The research methodology consisted of four main components; (1) collection tomato images of 3 sample species (holland tomato, sida tomato, and beef tomato) on three sides (topside, a slanted side, and a straight side), (2) image segmentation, (3) feature extraction about compactness, perimeter, area, radius, and standard deviation of radius, and (4) a tomato species classification using CNN algorithm. The tomato image collection and its characteristics (1,800 images for all samples) develop a CNN classification method. As a result, the selected variables to classify Holland tomato are perimeter and area (95.65% accuracy). The selected variables to classify Sida tomato classify by compactness and radius (86.52% accuracy). The selected variables to classify Beef tomato are Radius and standard deviation of radius using to classify Beef tomato (95.14% accuracy).