OPTIMASI DETEKSI WAJAH MENGGUNAKAN SUPER RESOLUTION DAN YOLO-SAHI

PATRA, FAUZAN ABGHI and Fachrurrozi, Muhammad Fachrurrozi and Rachmatullah, Muhammad Naufal (2025) OPTIMASI DETEKSI WAJAH MENGGUNAKAN SUPER RESOLUTION DAN YOLO-SAHI. Undergraduate thesis, Sriwijaya University.

[thumbnail of RAMA_55201_09021382126139_cover.pdf] Image
RAMA_55201_09021382126139_cover.pdf - Cover Image
Available under License Creative Commons Public Domain Dedication.

Download (149kB)
[thumbnail of RAMA_55201_09021382126139.pdf] Text
RAMA_55201_09021382126139.pdf - Accepted Version
Restricted to Repository staff only
Available under License Creative Commons Public Domain Dedication.

Download (3MB) | Request a copy
[thumbnail of RAMA_55201_09021382126139_TURNITIN.pdf] Text
RAMA_55201_09021382126139_TURNITIN.pdf - Accepted Version
Restricted to Repository staff only
Available under License Creative Commons Public Domain Dedication.

Download (9MB) | Request a copy
[thumbnail of RAMA_55201_09021382126139_0222058001_0001129204_01_front_ref.pdf] Text
RAMA_55201_09021382126139_0222058001_0001129204_01_front_ref.pdf - Accepted Version
Available under License Creative Commons Public Domain Dedication.

Download (1MB)
[thumbnail of RAMA_55201_09021382126139_0222058001_0001129204_02.pdf] Text
RAMA_55201_09021382126139_0222058001_0001129204_02.pdf - Accepted Version
Restricted to Repository staff only
Available under License Creative Commons Public Domain Dedication.

Download (398kB) | Request a copy
[thumbnail of RAMA_55201_09021382126139_0222058001_0001129204_03.pdf] Text
RAMA_55201_09021382126139_0222058001_0001129204_03.pdf - Accepted Version
Restricted to Repository staff only
Available under License Creative Commons Public Domain Dedication.

Download (457kB) | Request a copy
[thumbnail of RAMA_55201_09021382126139_0222058001_0001129204_04.pdf] Text
RAMA_55201_09021382126139_0222058001_0001129204_04.pdf - Accepted Version
Restricted to Repository staff only
Available under License Creative Commons Public Domain Dedication.

Download (689kB) | Request a copy
[thumbnail of RAMA_55201_09021382126139_0222058001_0001129204_05.pdf] Text
RAMA_55201_09021382126139_0222058001_0001129204_05.pdf - Accepted Version
Restricted to Repository staff only
Available under License Creative Commons Public Domain Dedication.

Download (902kB) | Request a copy
[thumbnail of RAMA_55201_09021382126139_0222058001_0001129204_06.pdf] Text
RAMA_55201_09021382126139_0222058001_0001129204_06.pdf - Accepted Version
Restricted to Repository staff only
Available under License Creative Commons Public Domain Dedication.

Download (162kB) | Request a copy
[thumbnail of RAMA_55201_09021382126139_0222058001_0001129204_07_ref.pdf] Text
RAMA_55201_09021382126139_0222058001_0001129204_07_ref.pdf - Bibliography
Restricted to Repository staff only
Available under License Creative Commons Public Domain Dedication.

Download (200kB) | Request a copy
[thumbnail of RAMA_55201_09021382126139_0222058001_0001129204_08_lamp.pdf] Text
RAMA_55201_09021382126139_0222058001_0001129204_08_lamp.pdf - Accepted Version
Restricted to Repository staff only
Available under License Creative Commons Public Domain Dedication.

Download (174kB) | Request a copy

Abstract

Face detection in low-resolution images remains a significant challenge in computer vision applications, particularly for surveillance systems and crowd analysis. This study proposes an integrated approach combining Enhanced Super-Resolution GAN (ESRGAN) for image quality enhancement and YOLO-SAHI (You Only Look Once with Slicing Aided Hyper Inference) for improved detection of small faces. The research utilizes the WIDER FACE dataset with varying difficulty levels to evaluate performance across different scenarios. Experimental results demonstrate that the hybrid approach significantly improves detection accuracy, with SAHI integration increasing recall by 35% for small faces compared to standalone YOLOv8. The complete YOLO-SAHI-ESRGAN system achieves 78.2% mean Average Precision (mAP) while maintaining reasonable processing latency of 420 ms per image. ESRGAN effectively enhances image quality, achieving PSNR of 24.43 dB and SSIM of 0.76 for easy cases. The study also develops a practical web-based implementation for real-world application.

Item Type: Thesis (Undergraduate)
Uncontrolled Keywords: Face detection, super-resolution, ESRGAN, YOLOv8, SAHI, mAP, PSNR
Subjects: T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK1-9971 Electrical engineering. Electronics. Nuclear engineering > TK1 Electrical engineering--Periodicals. Automatic control--Periodicals. Computer science--Periodicals. Information technology--Periodicals. Automatic control. Computer science. Electrical engineering. Information technology.
Divisions: 09-Faculty of Computer Science > 55201-Informatics (S1)
Depositing User: Fauzan abghi Patra
Date Deposited: 23 Sep 2025 06:49
Last Modified: 23 Sep 2025 06:49
URI: http://repository.unsri.ac.id/id/eprint/184647

Actions (login required)

View Item View Item