Proses absensi siswa secara manual selama ini masih menghadapi kendala, seperti memakan waktu lama, rentan salah catat, serta membuka celah kecurangan seperti titip absen. Berdasarkan observasi di SMK Karya Nasional Kuningan, pencatatan manual menghabiskan waktu sekitar 10 hingga 15 menit, yang akhirnya mengurangi efektivitas jam pelajaran. Menanggapi masalah tersebut, penelitian ini bertujuan merancang sistem absensi berbasis deteksi wajah menggunakan kolaborasi algoritma YOLOv8 dan FaceNet demi mendukung monitoring kehadiran secara real-time. Sistem ini dikembangkan dengan metode Waterfall, yang mencakup tahapan analisis kebutuhan, perancangan, implementasi, pengujian, hingga pemeliharaan. Perangkat lunak ini dibangun menggunakan bahasa pemrograman Python dengan memanfaatkan pustaka OpenCV, Flask, YOLOv8 untuk mendeteksi area wajah, FaceNet untuk pengenalan identitas, serta Firebase Realtime Database sebagai media penyimpanan data. Mekanisme kerja sistem dimulai dari kamera smartphone yang menangkap wajah siswa secara langsung. Algoritma YOLOv8 akan melokalisasi posisi wajah, kemudian FaceNet mencocokkan identitasnya lewat perbandingan embedding vector dengan dataset wajah yang sudah terdaftar. Begitu wajah dikenali, status kehadiran siswa otomatis tersimpan ke Firebase dan langsung tersinkronisasi ke dashboard admin. Pengujian sistem melibatkan 15 orang siswa dengan menguji beberapa skenario, seperti variasi pencahayaan, posisi wajah, dan jarak kamera. Hasil pengujian menunjukkan sistem dapat mengenali wajah secara optimal pada kondisi cahaya terang, posisi wajah lurus menghadap kamera, dan jarak maksimal 50 cm. Implementasi alat ini terbukti mempercepat pencatatan absen, menekan potensi titip absen, dan memudahkan sekolah dalam memantau kehadiran siswa. Kendati demikian, sistem masih memiliki keterbatasan pada lingkungan minim cahaya, wajah yang tertutup masker, atau kemiringan sudut wajah yang terlalu ekstrem. Secara keseluruhan, penelitian ini menyimpulkan bahwa penerapan algoritma YOLOv8 dan FaceNet mampu meningkatkan efisiensi waktu absensi, mempermudah pengawasan kehadiran secara real-time, sekaligus mendukung program digitalisasi sistem di lingkungan sekolah.

The manual student attendance process still faces several problems, such as taking a long time, being prone to recording errors, and allowing cheating practices such as attendance fraud. Based on observations conducted at SMK Karya Nasional Kuningan, the manual attendance recording process takes approximately 10 to 15 minutes, which reduces the effectiveness of learning time. In response to these problems, this study aims to design a face recognition-based attendance system using a combination of YOLOv8 and FaceNet algorithms to support real-time attendance monitoring. The system was developed using the Waterfall method, which includes the stages of requirement analysis, system design, implementation, testing, and maintenance. The software was built using the Python programming language by utilizing OpenCV, Flask, YOLOv8 for face detection, FaceNet for face recognition, and Firebase Realtime Database as the attendance data storage medium. The system mechanism begins with a smartphone camera capturing the student’s face in real-time. The YOLOv8 algorithm is used to localize the face position, while FaceNet identifies the face by comparing the embedding vector with the registered face dataset. Once the face is successfully recognized, the student attendance data is automatically stored in Firebase and synchronized directly with the admin dashboard. System testing involved 15 students using several testing scenarios, including variations in lighting conditions, face positions, and camera distances. The test results showed that the system was able to recognize faces optimally under adequate lighting conditions, with the face directly facing the camera, and at a maximum distance of 50 cm. The implementation of this system proved effective in accelerating the attendance recording process, reducing attendance fraud practices, and assisting the school in monitoring student attendance more efficiently. However, the system still has limitations under low-light conditions, when users wear masks, or when the face angle is excessively tilted. Overall, this study concludes that the implementation of YOLOv8 and FaceNet algorithms can improve attendance efficiency, facilitate real-time attendance monitoring, and support the digitalization of attendance systems in the school environment.