Earlier modes of attending roll call in learning institutes include manual roll call as well as roll calls by using cards; these methods are recessive to mistakes and proxy attendance. This paper focuses at designing an Automated Student Attendance System for facial recognition using CNN and Siamese Network. It lets the students login through a web interface, enter their roll number and, record 50 images for the training purpose. The used images are then normalized before being cropped to 64 × 64 to be used in training the model. These images are taken using a camera installed in the classroom and recorded in a particular excel sheet by marking a student as absent or present when before the start of each class. The effectiveness of the implementation is determined by the accuracy and matrices of confusion and by comparing the Siamese Network with a CNN model for real-time student recognition, the latter being superior. This way of work also lowers the impact of a human factor and increases the likelihood of fraudulent attendance.