DETEKSI FILTER CITRA WAJAH MENGGUNAKAN SUPPORT VECTOR MACHINE (SVM)

Wahyudi, Mohammad Irfan (2026) DETEKSI FILTER CITRA WAJAH MENGGUNAKAN SUPPORT VECTOR MACHINE (SVM). Undergraduate thesis, Universitas Nusantara PGRI Kediri.

[img] Text (Full Text)
RAMA_55201_2213020108.pdf - Accepted Version
Restricted to Registered users only
Available under License Creative Commons Public Domain Dedication.

Download (4MB) | Request a copy
[img] Text (Similarity)
RAMA_55201_2213020108_SIMILARITY.pdf - Accepted Version
Available under License Creative Commons Public Domain Dedication.

Download (964kB)
[img] Text (Cover sd BAB 1 + References)
RAMA_55201_2213020108_0703018704_0704108701_01_front_ref.pdf - Accepted Version
Available under License Creative Commons Public Domain Dedication.

Download (1MB)
[img] Text (BAB 2)
RAMA_55201_2213020108_0703018704_0704108701_02.pdf - Accepted Version
Restricted to Registered users only
Available under License Creative Commons Public Domain Dedication.

Download (566kB) | Request a copy
[img] Text (BAB 3)
RAMA_55201_2213020108_0703018704_0704108701_03.pdf - Accepted Version
Restricted to Registered users only
Available under License Creative Commons Public Domain Dedication.

Download (751kB) | Request a copy
[img] Text (BAB 4)
RAMA_55201_2213020108_0703018704_0704108701_04.pdf - Accepted Version
Restricted to Registered users only
Available under License Creative Commons Public Domain Dedication.

Download (565kB) | Request a copy
[img] Text (BAB 5)
RAMA_55201_2213020108_0703018704_0704108701_05.pdf - Accepted Version
Restricted to Registered users only
Available under License Creative Commons Public Domain Dedication.

Download (218kB) | Request a copy
[img] Text (References)
RAMA_55201_2213020108_0703018704_0704108701_06_ref.pdf - Bibliography
Available under License Creative Commons Public Domain Dedication.

Download (207kB)
[img] Text (Lampiran)
RAMA_55201_2213020108_0703018704_0704108701_07_lamp.pdf - Accepted Version
Restricted to Registered users only
Available under License Creative Commons Public Domain Dedication.

Download (2MB) | Request a copy

Abstract

Penggunaan beauty filter di media sosial dapat menurunkan keandalan sistem pengenalan wajah otomatis karena memanipulasi karakteristik seperti tekstur, gradien, dan kontur asli wajah. Penelitian ini bertujuan merancang model sistem pendeteksi citra wajah kategori "Asli" dan "Filter". Sistem deteksi dibangun menggunakan algoritma Support Vector Machine (SVM) sebagai klasifikasi utama. Proses diawali prapengolahan lokalisasi area wajah menggunakan Haar Cascade, dilanjutkan dengan kombinasi ekstraksi fitur Histogram of Oriented Gradients (HOG) untuk pola gradien geometri wajah dan Local Binary Pattern (LBP) untuk detail tekstur kulit. Dataset penelitian menggunakan data primer berformat citra digital (.jpg, .jpeg, .png) dengan batasan efek/filter kecantikan (beauty filter) dari media sosial. Pengujian kinerja model klasifikasi SVM dievaluasi secara kuantitatif melalui Confusion Matrix berdasarkan metrik Akurasi, Presisi, Recall, dan F1-score. Penelitian ini menggunakan uji coba 3 kernel, Radial Basis Function (RBF), linear, dan polynomial. Hasil uji coba kernel RBF dan linear memperoleh akurasi 85%, Selain itu, kedua kernel tersebut memperoleh nilai precision, recall, dan F1-score yang lebih baik dibandingkan kernel Polynomial, terutama dalam mendeteksi citra wajah asli dan citra wajah berfilter. Kernel Polynomial juga mampu memberikan performa klasifikasi yang cukup baik dengan nilai akurasi sebesar 82%.

Item Type: Thesis (Undergraduate)
Uncontrolled Keywords: Deteksi Filter Wajah, Support Vector Machine (SVM), Histogram of Oriented Gradients (HOG), Local Binary Pattern (LBP), Haar Cascade
Subjects: 410 Engineering science > 457 Computer engineering
410 Engineering science > 459 Computer science
710 Education science > 786 Informatics engineering
Divisions: Fakultas Teknik dan Ilmu Komputer > S1-Teknik Informatika
Depositing User: Mohammad Irfan Wahyudi
Last Modified: 06 Aug 2026 01:37
URI: http://repository.unpkediri.ac.id/id/eprint/27090

Actions (login required)

View Item View Item