This paper presents the detailed implementation and application of the new attendance management system in academic libraries through the use of deep learning face-recognition technology and computer vision to address the drawbacks of traditional attendance check methods. The main idea of the system relies on a well-experienced module through the use of machine learning, a pre-trained model, and a database that contributes to the system's ability to identify the attendees and log in their names, identifications, dates, and times. The study relied on an experimental approach to help determine the extent of the ability of the proposed system to register beneficiaries' entry into the library efficiently and accurately without any problems occurring. The findings and experimental results show that the proposed system is accurate, fast, reliable and able to recognize up to four faces simultaneously without any technical issues. From the results of the proposed system's accuracy test, it was found that the accuracy of attendance checks when recognizing only one face of a beneficiary was 100%. The accuracy of the attendance checks when recognizing the beneficiary with or without a cap on the head was 100%. The accuracy of attendance checks when recognizing two faces at one time was 100%. The accuracy of attendance checks when recognizing four faces together at one time was 100%. The accuracy of the attendance checks when the beneficiary is facing forward was 100%. The attendance checks' accuracy when the beneficiary faces sideways is 95%.