Abstract
Pencatatan kehadiran mahasiswa adalah elemen penting terutama dalam pembelajaran berbasis proyek (Project Based Learning) yang memerlukan kolaborasi dan partisipasi aktif. Penelitian ini mengembangkan sistem absensi otomatis berbasis face recognition dengan menggunakan model Multi-task Cascaded Convolutional Neural Network (MTCNN) untuk deteksi wajah. Model MTCNN dipilih karena kemampuannya dalam mendeteksi wajah dengan baik meski terdapat variasi sudut, pencahayaan, dan ekspresi. Proses deteksi melibatkan tiga tahap seperti deteksi wajah, penghalusan kotak pembatas (bounding box) dan penyesuaian titik penanda (landmark). Setelah itu metode Eigenface digunakan untuk mengenali wajah dengan analisis Principal Component Analysis (PCA). Penelitian diuji di kelas yang sebennarnya dengan variasi jumlah mahasiswa, pencahayaan, dan posisi kamera. Hasilnya metode MTCNN mendeteksi wajah dengan rata-rata 29.8% presisi, sementara Eigenface mencapai akurasi 90%, meski turun menjadi 78% untuk subjek berkacamata. Selain itu MTCNN menunjukkan waktu deteksi lebih cepat dibanding RetinaFace yang menandakan efisiensi lebih tinggi. Kombinasi MTCNN dan Eigenface terbukti efektif dalam otomatisasi pencatatan kehadiran dengan peningkatan efisiensi dan akurasi data yang mendukung pelaksanaan pembelajaran berbasis proyek secara optimal.
Kata kunci : Face Recognition, MTCNN, Eigenface, Project-Based Learning
Student attendance tracking is a crucial element, particularly in Project-Based Learning (PBL) which requires collaboration and active participation. This research develops an automatic attendance system based on face recognition using the Multi-task Cascaded Convolutional Neural Network (MTCNN) model for face detection. MTCNN was selected for its robust ability to detect faces despite variations in angle, lighting, and expression. The detection process involves three stages: face detection, bounding box refinement, and landmark adjustment. Following detection, the Eigenface method is used for face recognition with Principal Component Analysis (PCA). The research was tested in an actual classroom with varying student numbers, lighting conditions, and camera positions. Results showed that MTCNN achieved an average face detection precision of 29.8%, while Eigenface reached 90% accuracy, though it dropped to 78% for subjects with glasses. Additionally, MTCNN demonstrated faster detection times compared to RetinaFace, indicating higher efficiency. The combination of MTCNN and Eigenface proved effective in automating attendance recording, enhancing data accuracy and efficiency to optimally support Project-Based Learning.
Keywords: Face Recognition, MTCNN, Eigenface, Project-Based Learning