Jurnal Mantik
Vol. 5 No. 4 (2022): February: Manajemen, Teknologi Informatika dan Komunikasi (Mantik)

Implementation of Random Forest for Motif Classification Based on Sift

Ahkyar Khadafi (Universitas Pembangunan Panca Budi)
Muhammad Iqbal (Universitas Pembangunan Panca Budi)



Article Info

Publish Date
25 Feb 2022

Abstract

Songket an example of intangible cultural heritage found in Indonesia. Songket, especially Songket Palembang, has a lot of variety depending on the features of each place. When compared to Songket from other locations, Palembang Songket has more features. Songket Palembang has a high motive, quality, and complexity in the manufacturing process, in addition to its historical significance. Using Scale-Invariant Feature Transform (SIFT) feature extraction, the Random Forest approach was employed to identify the Songket Palembang motif image in this study. The SIFT approach involves the steps of extrema detection scale space, keypoint localization, orientation assignment, and keypoint descriptor in the feature creation process. The Random Forest categorization uses the resulting feature. In this study, 115 photographs of each sort of Songket theme were used, including Chinese Flowers, Beautiful Flowers, and Pulir. Each Songket Palembang motif's five colors were used to create the image. For each Songket Palembang motif, 100 and 15 training and test data were used, respectively. The test results show that the SIFT and Random Forest methods for the classification of Songket Palembang motifs can provide fairly good accuracy, with overall accuracy of 92.98 percent, per class accuracy of 94.07 percent, precision 92.98 percent, and recall 89.74 percent for the SIFT and Random Forest methods, respectively.

Copyrights © 2022






Journal Info

Abbrev

mantik

Publisher

Subject

Computer Science & IT Economics, Econometrics & Finance Languange, Linguistic, Communication & Media

Description

Jurnal Mantik (Manajemen, Teknologi Informatika dan Komunikasi) is a scientific journal in information systems/informati containing the scientific literature on studies of pure and applied research in information systems/information technology,Comptuer Science and management science and public ...