Identifikasi Wajah Tersamar berdasarkan Data Pelatihan Wajah Tak Sempurna Menggunakan Pendekatan Deep Learning

Dinata, Rangga Kusuma (2021) Identifikasi Wajah Tersamar berdasarkan Data Pelatihan Wajah Tak Sempurna Menggunakan Pendekatan Deep Learning. Undergraduate thesis, Institut Teknologi Sepuluh Nopember.

[img] Text
05111740000120-Undergraduate_Thesis.pdf - Accepted Version
Restricted to Repository staff only until 1 October 2023.

Download (5MB) | Request a copy
[img] Text
05111740000120-Undergraduate_Thesis.pdf
Restricted to Repository staff only

Download (5MB) | Request a copy
[img] Text
05111740000120-Undergraduate_Thesis.pdf
Restricted to Repository staff only

Download (5MB) | Request a copy

Abstract

Sistem pengenalan wajah merupakan teknologi yang dapat melakukan proses identifikasi atau verifikasi seseorang dari citra digital atau video frame dari sumber video. Perubahan wajah dapat secara dramatis menyamarkan identitas seseorang dengan memasukkan beragam atribut fisik yang diubah seperti mengenakan topi atau masker wajah, mengubah gaya rambut, atau warna rambut, mengenakan kacamata hitam, mencukur atau menumbuhkan jenggot. Dengan perkembangan teknologi saat ini terutama akselerasi transformasi digital selama pandemi COVID-19, kebutuhan sistem pengenalan wajah dituntut untuk lebih canggih dalam mengenali wajah tersamar. Dalam tugas akhir ini, dilakukan pengembangan identifikasi wajah tersamar berdasarkan data pelatihan wajah tak sempurna menggunakan pendekatan deep learning. Dataset yang digunakan adalah dataset yang diolah secara mandiri dari data rekaman video responden dan dataset citra Brazilian FEI. Pengujian dilakukan dengan menggunakan data pengujian yang menghasilkan hasil berupa prediksi kelas, nilai akurasi, F1-score, recall dan precision. Evaluasi dilakukan pada data yang diolah secara mandiri yang memiliki 17 subjek kelas. Hasil dari tugas akhir ini menunjukkan bahwa performa evaluasi terhadap identifikasi wajah tersamar setiap frame video dengan menggunakan Support Vector Machine, model yang dibangun menghasilkan akurasi sebesar 87.05%, rata-rata makro F1-score sebesar 85.52%, rata-rata makro recall sebesar 87.24%, dan rata-rata makro precision sebesar 91.45%. Identifikasi wajah setiap video menghasilkan akurasi sebesar 94.74%, rata-rata makro F1-score sebesar 92.16%, rata-rata makro recall sebesar 94.12%, dan rata-rata makro precision sebesar 91.18%. ================================================================================================================== A facial recognition system is a technology that enables personal identification or verification through digital images or frames from a video. Dramatic facial changes could disguise someone's identity by adding changeable various physical attributes like wearing a cap or face mask, changing hairstyle or hair colour, wearing sunglasses, trimming or growing a beard. Given technology development nowadays, especially the acceleration of digital transformation during the COVID-19 pandemic, the need for facial recognition systems demanded to be more sophisticated for recognizing a disguised face. In this final project, disguised face identification is developed based on imperfect facial data using a deep learning approach. The datasets that being used are datasets that had been curated independently from respondent's video recording and Brazilian FEI images dataset. Testing is carried out by using testing data that yield results like class prediction, accuracy, F1-score, recall and precision. Evaluation is carried out in the curated datasets from the respondent's video recording with 17 subject classes. The results of this final project indicate that the best performance evaluation toward disguised face identification per video frame using Support Vector Machine, a developed model yields an accuracy of 87.05%, macro-average F1-score of 85.52%, macro-average recall of 87.24% and macro-average precision of 91.45%. Face identification per video yields an accuracy of 94.74%, macro-average F1-score of 92.16%, macro-average recall of 94.12% and macro-average precision of 91.18%.

Item Type: Thesis (Undergraduate)
Uncontrolled Keywords: Identifikasi Wajah Tersamar, Data Wajah Tak Sempurna, Deep Learning, Disguised Face Identification, Imperfect Facial Data
Subjects: Q Science > Q Science (General) > Q325.5 Machine learning.
Q Science > QA Mathematics > QA336 Artificial Intelligence
R Medicine > R Medicine (General) > R858 Deep Learning
T Technology > TA Engineering (General). Civil engineering (General) > TA1650 Face recognition. Optical pattern recognition.
Divisions: Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User: Rangga Kusuma Dinata
Date Deposited: 19 Aug 2021 07:31
Last Modified: 19 Aug 2021 07:31
URI: https://repository.its.ac.id/id/eprint/87681

Actions (login required)

View Item View Item