Attendance tracking is a tedious task, especially in educational institutions with a multitude of students. Traditional attendance methods like roll calls and RFID systems are time-consuming, prone to errors, and lack robust security, leading to inefficiencies in large-scale applications. Several face recognition-based attendance logging methods have been reported in the literature so far. But most of the approaches are prone to false positives, and lack robustness in varying conditions and struggles with face alignment. In this study, a new approach to image-based attendance logging system is introduced, leveraging multi-task convolutional neural network (MTCNN) which not only provides accurate face alignment and recognition, but also detects and recognizes the faces properly. Extensive simulation studies demonstrate that the proposed approach attains an accuracy of up to 97% in attendance logging compared to other methods.