TY  - THES
N1  - Nurochman, S.Kom., M.Kom.
ID  - digilib77704
UR  - https://digilib.uin-suka.ac.id/id/eprint/77704/
A1  - Ahmad Bidni Musyafa, NIM.: 22106050063
Y1  - 2026/06/05/
N2  - Manual security monitoring in female-only dormitories has limitations and is prone to negligence. To address this issue, this study designs a gender detection system as a proactive early warning system. The system utilizes deep learning-based computer vision technology to perform automatic and Real-time security monitoring.
The system was developed using a full-body recognition approach with the lightweight YOLOv8n architecture, integrated with automatic notifications via Telegram. The dataset consists of simulated CCTV images (semi top-down angles, lighting variations, and mask/helmet occlusion scenarios) as well as secondary data. The entire data was divided into 70% training data, 20% validation data, and 10% testing data using a transfer learning strategy.
Evaluation results show the model achieved an mAP50 of 92.90%, precision of 90.65%, recall of 89.01%, and F1-score of 89.82%. The system proved efficient, operating stably at an average speed of 20 FPS on purely CPU-based devices without GPU acceleration. Field testing demonstrated that the full-body recognition approach remains robust in detecting subjects under occlusion. In conclusion, the YOLOv8n model provides an optimal balance between high accuracy, local computational efficiency, and operational reliability.
PB  - UIN SUNAN KALIJAGA YOGYAKARTA
KW  - Yolov8n
KW  -  Deteksi Gender
KW  -  Full-Body Recognition
KW  -  Computer Vision
KW  -  Sistem Real-Time
KW  -  Early Warning System
M1  - skripsi
TI  - RANCANG BANGUN SISTEM DETEKSI GENDER MENGGUNAKAN MODEL YOLOV8 UNTUK PEMANTAUAN KEAMANAN PADA HUNIAN KHUSUS PUTRI BERBASIS FULL-BODY RECOGNITION
AV  - restricted
EP  - 127
ER  -