Analisis Performa Sistem Pendeteksi Wajah Berbasis Convolutional Neural Network (CNN)

Najwa, Rizky (2019) Analisis Performa Sistem Pendeteksi Wajah Berbasis Convolutional Neural Network (CNN). Other thesis, Insitut Teknologi Sepuluh Nopember.

Text 07211540000039-Undergraduate_Thesis.pdf Restricted to Repository staff only Download (16MB) | Request a copy

Abstract

Pengenalan wajah merupakan sebuah proses pengambilan citra atau data dari wajah seseorang dan kemudian menghubungkan data tersebut dengan seorang individu yang spesifik. Karena wajah merupakan salah satu karakteristik fisik manusia sulit untuk dipindah tangankan, maka sistem pengenalan wajah sering diaplikasikan ke dalam sistem autentikasi identitas dan sistem keamanan yang berbasis biometrik. Namun, masih terdapat beberapa tantangan dalam pengenalan wajah, yang menyebabkan pengenalan wajah sulit dilakukan. Sistem pengenalan wajah terdiri dari tiga proses utama yaitu deteksi wajah, ekstraksi fitur, dan pengenalan wajah. Pada tugas akhir ini akan dianalisis performa pada tahap deteksi wajah yang dibuat dengan menggunakan sistem berbasis Convolutional Neural Network (CNN), yaitu Single Shot Multibox Detector (SSD) dan You Only Look Once (YOLO). Pengujian dilakukan dengan melakukan perekaman skenario kondisi wajah yang kemudian dilewatkan ke sistem berbasis CNN dan dianalisis dengan menggunakan nilai mean Average Precision (mAP). Berdasarkan hasil pengujian, sistem berbasis SSD mampu melakukan pendeteksian pada skenario pengujian di siang hari dengan rata-rata waktu pemrosesan selama 8.15 detik, rata-rata FPS sebesar 25.82 fps, dan rata-rata nilai mAP sebesar 48.11%. Pada malam hari sistem berbasis SSD membutuhkan rata-rata waktu pemrosesan selama 7.92 detik, rata-rata FPS sebesar 26.4 fps, dan rata-rata nilai mAP sebesar 34.69%. Sedangkan sistem berbasis YOLO mampu melakukan pendeteksian pada skenario pengujian di siang hari dengan rata-rata waktu pemrosesan selama 30.49 detik, rata-rata FPS sebesar 6.89 fps, dan rata-rata nilai mAP sebesar 60.65%. Pada malam hari sistem berbasis YOLO membutuhkan rata-rata waktu pemrosesan selama 30.48 detik, rata-rata FPS sebesar 6.82 fps, dan rata-rata nilai mAP sebesar 71.65%.
============================================================================================================================================
Face recognition is a process of taking an image or visual data from someone's face and then connecting that data with a specific individual. Because faces are one of the physical characteristics of humans that are difficult to transfer, the face recognition system is often applied to identity authentication systems and biometric-based security systems. However, there are still some challenges in face recognition, which makes face recognition difficult. The face recognition system consists of three main parts, namely face detection, feature extraction, and facial recognition. In this final project, the performance of the face detection stage will be analyzed using Convolutional Neural Network (CNN) based systems, namely Single Shot Multibox Detector (SSD) and You Only Look Once (YOLO). Testing is done by recording face condition scenarios which are then passed to CNN-based systems and analyzed using mean Average Precision (mAP) value. Based on the test results, the SSD-based system is able to detect the test scenario during the day with an average processing time of 8.15 seconds, the average FPS is 25.82 fps, and the average mAP value is 48.11%. At night the SSD-based system requires an average processing time of 7.92 seconds, the average FPS is 26.4 fps, and the average mAP value is 34.69%. While the YOLO-based system is able to detect the test scenario during the day with an average processing time of 30.49 seconds, the average FPS is 6.89 fps, and the average mAP value is 60.65%. At night the YOLO-based system requires an average processing time of 30.48 seconds, the average FPS is 6.82 fps, and the average mAP value is 71.65%.

Item Type:	Thesis (Other)
Additional Information:	RSKom 006.42 Naj a-1 2019
Uncontrolled Keywords:	System Performance Analysis, Face Detection System, Single Shot Multibox Detector (SSD), You Only Look Once (YOLO), Convolutional Neural Network (CNN).
Subjects:	Q Science > QA Mathematics > QA76.6 Computer programming. Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Divisions:	Faculty of Electrical Technology > Computer Engineering > 90243-(S1) Undergraduate Thesis
Depositing User:	Najwa Rizky
Date Deposited:	06 Aug 2026 04:26
Last Modified:	06 Aug 2026 04:26
URI:	http://repository.its.ac.id/id/eprint/70447

Actions (login required)

View Item