Sagala, Jody Hezekiah Tanasa (2026) Pengembangan Sistem Monitoring Presensi Berbasis Re-Identification (OSNet) dan Face Recognition (ArcFace) Untuk Validasi Kehadiran Mahasiswa Secara Real-Time. Other thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 5027221050-Undergraduate_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
5027221050-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (10MB) | Request a copy |
Abstract
Presensi mahasiswa merupakan parameter fundamental dalam administrasi akademik untuk menjamin validitas proses pembelajaran. Namun, implementasi pengenalan wajah standar sering kali terhambat oleh kendala oklusi spasial, variasi sudut pandang ekstrem, dan keterbatasan sumber daya komputasi pada edge device. Penelitian ini mengembangkan sistem monitoring presensi berkelanjutan yang mengintegrasikan deteksi objek hibrida YOLOv8n, pelacak BoTSORT, dan pangkalan data vektor in-memory USearch berbasis algoritma Hierarchical Navigable Small World (HNSW). Untuk memitigasi Identity Switch saat wajah teroklusi atau subjek membelakangi kamera, diimplementasikan mekanisme kontingensi berjenjang (Cascaded Fallback Re-ID) berbasis ekstraksi fitur bodi OSNet-AIN. Hasil pengujian stress-testing 1.000 iterasi terhadap 22 subjek menunjukkan akurasi True Positive Rate (TPR) global sebesar 96,50% dengan False Acceptance Rate (FAR) mencapai 0,00%. Evaluasi operasional pada kondisi riil membuktikan efisiensi komputasi sistem dengan rata-rata waktu respons identifikasi instan (Time-to-First-Match) sebesar 927,53 ms pada kondisi ideal, serta mekanisme fallback yang tetap menjaga integritas presensi pada kondisi oklusi natural hingga rata-rata 3.571,10 ms. Dengan stabilitas frame rate operasional pada 50 FPS dan rata-rata latensi kueri vektor hanya 0,42 ms, sistem terbukti sukses menjamin persistensi log presensi yang objektif dan real-time tanpa ketergantungan pada jaringan eksternal.
======================================================================================================================================
Student attendance recording is a fundamental parameter in academic administration to ensure the validity of the learning process. However, conventional attendance systems and standard face recognition implementations are often hindered by spatial occlusions, extreme viewpoint variations, and computational resource constraints on edge devices. This research develops a continuous attendance monitoring system that integrates a hybrid YOLOv8n object detection model, the BoTSort tracker, and an in-memory vector database using the Hierarchical Navigable Small World (HNSW) algorithm via the USearch library. To mitigate Identity Switch errors when faces are occluded or subjects are facing away from the camera, a Cascaded Fallback Re-Identification (Re-ID) mechanism based on OSNet-AIN body feature extraction is implemented. Stress-testing results from 1,000 iterations involving 22 subjects demonstrate a global True Positive Rate (TPR) of 96.50% with a False Acceptance Rate (FAR) of 0.00%. Real-world operational evaluations confirm the system's computational efficiency, achieving an average Time-to-First-Match (TTFM) of 927.53 ms under ideal conditions, and a resilient fallback mechanism that maintains attendance log integrity during natural occlusions with an average latency of 3,571.10 ms. With a stable operational frame rate of 50 FPS and an average vector search latency of only 0.42 ms, the system successfully guarantees objective, real-time attendance logging without reliance on external network connectivity.
| Item Type: | Thesis (Other) |
|---|---|
| Uncontrolled Keywords: | Monitoring Presensi, Multi-Object Tracking, Person Re-Identification, YOLOv8, BoTSORT, USearch, OSNet, ArcFace, Attendance Monitoring |
| Subjects: | T Technology > TA Engineering (General). Civil engineering (General) > TA1650 Face recognition. Optical pattern recognition. |
| Divisions: | Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information Technology > 59201-(S1) Undergraduate Thesis |
| Depositing User: | Jody Hezekiah Tanasa Sagala |
| Date Deposited: | 04 Aug 2026 01:47 |
| Last Modified: | 04 Aug 2026 01:47 |
| URI: | http://repository.its.ac.id/id/eprint/142732 |
Actions (login required)
 |
View Item |