Checking attendance in a classroom is a factor contributing to the final performance of the students in the course. For both students and professors, attendance checking by name is very time-consuming and, in particular, the latter is very susceptible to simple attendance fraud. The study used a Face Recognition based attendance method using the YOLOv3 approach as an alternative. The system, based on face-detection and face-recognition algorithms, automatically recognizes, and marks attendance by recognizing the student. The experimental result shows that by using the trained model with a training accuracy of 98.01%, the proposed attendance system achieved 94% face recognition efficiency.