Work

Implementasi YOLOv8 untuk Absensi Siswa Berbasis Deteksi Wajah Menggunakan Webcam: Indonesia

Year: 2026
Type: article
Abstract: Abstrak. Sistem absensi manual di sekolah memiliki berbagai keterbatasan dalam hal efisiensi dan kerentanan terhadap kesalahan pencatatan maupun manipulasi data. Penelitian ini bertujuan mengimplementasikan algoritma YOLOv8 pada sistem absensi siswa berbasis deteksi wajah menggunakan webcam, serta menguji kinerja sistem tersebut dalam memproses dan mencatat kehadiran siswa secara real-time. Studi kasus dilakukan di SMA Swasta Assisi Siantar dengan populasi seluruh siswa dan sampel sebanyak 105 siswa. Metode penelitian bersifat eksperimen kuantitatif yang meliputi tahapan akuisisi citra wajah, anotasi, pra-pemrosesan, dan pelatihan model YOLOv8 dengan dua skenario, yaitu 50 dan 75 epoch. Hasil penelitian menunjukkan bahwa model dengan 75 epoch memberikan performa terbaik dengan mAP50 sebesar 0,9156, precision 0,9012, recall 0,8897, dan F1 Score 0,8954, lebih unggul dibandingkan model 50 epoch. Algoritma YOLOv8 berhasil diimplementasikan dan diintegrasikan ke dalam sistem absensi berbasis website, mampu mendeteksi wajah secara real-time melalui webcam dan mencocokkannya dengan data wajah pada basis data sebagai dasar pencatatan kehadiran otomatis. Pengujian terhadap seluruh 105 siswa menunjukkan tingkat keberhasilan pengenalan 100% dengan waktu pengenalan 1,7 hingga 2,0 detik pada model terbaik. Kinerja sistem dinilai fungsional dan responsif dalam menjalankan pra-pemrosesan secara end-to-end, mendeteksi wajah dengan tingkat keyakinan (confidence) yang memadai (mencapai 73,1%), serta secara otomatis mencatat kehadiran siswa kelas X dan XI ke dalam basis data beserta waktu kehadirannya tanpa kendala. Kata Kunci: Absensi; Pengenalan Wajah; YOLOv8; Real-time; Webcam. Abstract. Manual attendance systems in schools have various limitations in terms of efficiency and vulnerability to recording errors and data manipulation. This study aims to implement the YOLOv8 algorithm in a webcam-based, face-detection student attendance system and to evaluate its performance in processing and recording student attendance in real time. The case study was conducted at SMA Swasta Assisi Siantar, with the entire student population and a sample of 105 students. The research applies a quantitative experimental method covering face image acquisition, annotation, preprocessing, and YOLOv8 model training under two scenarios, namely 50 and 75 epochs. The results show that the 75-epoch model achieved the best performance, with an mAP50 of 0.9156, precision of 0.9012, recall of 0.8897, and an F1 score of 0.8954, outperforming the 50-epoch model. The YOLOv8 algorithm was successfully implemented and integrated into a website-based attendance system, capable of detecting faces in real time through a webcam and matching them against stored facial data as the basis for automatic attendance recording. Testing on all 105 students showed a 100% recognition success rate, with recognition times of 1.7 to 2.0 seconds for the best model. The system's performance is considered functional and responsive in executing end-to-end preprocessing, detecting faces with adequate confidence (reaching 73.1%), and automatically recording the attendance of grade X and XI students into the database along with their attendance time without any issues. Keywords: Attendance; Face Recognition; YOLOv8; Real-time; Webcam.
Language: English
Open Access status: diamond
OpenAlex ID: W7235813986
FWCI:
Cites: 17
Cited by:
Related to: 10