Pengembangan Aplikasi Pembelajaran Mesin untuk Identifikasi Kemiripan Lukisan

Abstract

This study has one goal, namely to be able to identify the similarity of paintings through painting imagery and from the image of the painting that has been input will show which artist's work is supposedly similar to the image that has been input. The algorithm used is a Convolutional Neural Network (CNN). This study used painting imagery data from Kaggle in the form of a folder. The image data used was 4299 which was divided into training data and testing data with a total of 3444 data from 11 classes and testing data as many as 855 from 11 classes. The framework used in this study is ResNet-50 and the Convolutional Neural Network which is applied is Tensorflow Keras. The results of the study, it has produced an accuracy value of 24% with an average probability of up to about 80% and above.