Implementasi Fitur Sift Dan Fisher Vector Encoding Untuk Aplikasi Pengenalan Wajah

Yudhantoro, Yusuf Azis Henny Tri (2017) Implementasi Fitur Sift Dan Fisher Vector Encoding Untuk Aplikasi Pengenalan Wajah. Undergraduate thesis, Institut Teknologi Sepuluh Nopember.

[img] Text
5112100086_Undergraduate_Theses.pdf - Published Version
Restricted to Repository staff only

Download (1MB)

Abstract

Teknik pengenalan individu berbasis fitur biometrik wajah menjadi salah satu teknik yang paling sering digunakan. Hal ini dikarenakan penggunaan wajah untuk pengenalan memiliki beberapa kelebihan, antara lain: pengenalan wajah hanya membutuhkan peralatan kamera yang relatif ekonomis dan algoritma yang baik mampu mengidentifikasi wajah. Dalam Tugas Akhir ini diimplementasikan perangkat lunak pengenalan wajah menggunakan Scalable Invariant Feature Transform (SIFT) mode dense, dan fisher vector encoding, serta mahalanobis metric learning. Hasil dari proses metric learning adalah threshold kesamaan wajah dan bobot jarak fisher vector pasangan gambar wajah yang berguna dalam proses klasifikasi. Dataset yang digunakan dalam proses uji coba berisi empat ratus pasangan gambar kelas positif dan negatif dari Labeled Faces in the Wild (LFW). Akurasi terbaik sebesar 53%. ================================================================================================================== Individual recognition techniques based on facial biometric features is one of the most frequently used technique. due to the usage of facial recognition has several advantages such as: facial recognition process only need relatively economic camera equipment and a good algorithm to be able to identify a face. In this undergraduate theses, facial recognition software using dense mode of Scalable Invariant Feature Transform (SIFT) and fisher vector encoding to extract features. Weight of pair-faceimage’s fisher vector and distance threshold to classify images learned using mahalanobis metric learning. The dataset used in the testing process contains four hundred pairs of positive and negative class picture from Labeled Faces in the Wild (LFW). The best accuration is 53%.

Item Type: Thesis (Undergraduate)
Uncontrolled Keywords: Biometric, Pengenalan Wajah, SIFT, Fisher Vector, Mahalanobis Metric Learning
Subjects: T Technology > T Technology (General)
T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques
Divisions: Faculty of Information Technology > Informatics Engineering > (S1) Undergraduate Theses
Depositing User: - YUSUF AZIS HENNY TRI YUDHANTORO
Date Deposited: 07 Apr 2017 03:33
Last Modified: 27 Dec 2017 07:32
URI: http://repository.its.ac.id/id/eprint/3168

Actions (login required)

View Item View Item