Chili classification automation using deep learning with convolutional neural network method (Case study: Three types of chili with form of curly, cayenne and binocular)

Abstract

Chili can be easily found in markets, minimarkets or large supermarkets, only the ones that are easily found are curly chili, cayenne pepper and binocular chili. Warehouse or storage is a place for storing goods both raw materials which are then processed or manufactured goods that are ready and marketed. CNN is a machine learning method in which the development of Multi-Layer Perceptron (MLP) is designed to process two-dimensional data. The purpose of this study is to classify the image of curly chili, cayenne and binoculars so that it becomes an input for warehousing to apply the CNN method in placing goods. Obtained the results of this study is the use of the best epoch value of 70 with 80%: 20% training testing dataset comparison scenarios and, obtained a classification accuracy of 70%, where, in the classification results of the predicted chili image according to the actual chili image only 10 image for curly chili and 7 image for cayenne image and 4 image for binocular chili image.