Litcius/Paper detail

InceptionV3, ResNet50, and VGG19 Performance Comparison on Tomato Ripeness Classification

John Raphael Pangilinan, Jericho Legaspi, Noel B. Linsangan

202229 citationsDOI

Abstract

Automated machine learning-based technology has become more critical in the agriculture and food processing sectors. Various algorithms are utilized in creating image processing classification systems. However, the performance of these algorithms is not often compared with one another, especially when applied to tomato ripeness classification. This study assessed the performance of three Convolutional Neural Network algorithms (InceptionV3, ResNet50, and VGG19) in terms of accuracy in identifying the ripeness of tomatoes. The experimental setup was composed of a 3D printed case that houses the Raspberry Pi 4 model B, an LED strip, and a Raspberry Camera Module v1.3 as its main hardware components. Sixty (60) tomato samples were utilized to test each algorithm, and a confusion matrix table was applied to the data gathered. The VGG19 algorithm garnered the highest accuracy score of 95%, followed by the ResNet50 algorithm with 93.33%, and lastly, the InceptionV3 algorithm with 91.67%.

Topics & Concepts

RipenessConfusion matrixComputer scienceArtificial intelligenceConvolutional neural networkMachine learningAlgorithmRipeningFood scienceChemistrySmart Agriculture and AISpectroscopy and Chemometric AnalysesLeaf Properties and Growth Measurement
InceptionV3, ResNet50, and VGG19 Performance Comparison on Tomato Ripeness Classification | Litcius