Sistem Absensi Berbasis Deteksi Wajah dengan Pendekatan Eksperimen

Meryatul Husna
Kinarta Ketaren
Sharfina Faza
Orli Binta Tumanggor
Aprilza Aswani

Jurnal Sains dan Teknologi
Vol. 4 No. 6 (2025): Desember 2025

This study compares the accuracy of two face detection algorithms, Haar Cascade and Multi-task Cascaded Convolutional Networks (MTCNN), to determine the most suitable method for implementation in a facial recognition–based attendance system. The evaluation was conducted through a series of tests under common real-world conditions, including variations in distance, lighting intensity, and face orientation. Each algorithm was assessed using performance metrics such as Precision, Recall, F1-Score, and processing time to provide a comprehensive understanding of their strengths and limitations. The results indicate that MTCNN consistently achieves higher accuracy across nearly all tested scenarios, particularly under low-light conditions and when the face is not oriented frontally. In contrast, Haar Cascade demonstrates faster processing time but experiences significant decreases in accuracy under non-ideal conditions typically found in practical applications. Based on these findings, MTCNN is considered more suitable for attendance systems that require high accuracy and robustness to environmental variations, while Haar Cascade may be preferred in applications where computational efficiency and speed are the primary considerations.