Automated Attendance System using Few-Shot Learning Approach

We are living in a world where artificial intelligence is slowly reshaping the world. Artificial Intelligence systems reduce the error rate and can work very efficiently and with incredible speed and precision. Nowadays, Artificial Intelligence is being used to make the technology smarter so it can take decisions and minimize human efforts.An “Automatic attendance system using facial recognition” is one of the smart implementations of Artificial Intelligence. When we stand in front of the camera, it will take our photo and match it with the photos it has in its database. Once the photo is matched with a photo from the database, the system marks that person's attendance as present in the table and stores it in the database so that we can view the attendance sheet as needed.In this paper we are mainly focusing on one of the major problems in making these systems i.e., requirement of large training data. As the machine learning and deep learning model require a large training data for its model to work with high accuracy, it become a tedious task to collect large no of photos per person. So, in the proposed model we are trying to minimize the training data up to 5 photos per person in a way that it doesn't affect the accuracy of the whole system. We are using Convolutional neural network and pretrained FaceNet model and we are getting an accuracy of approx. 98% with 100 different faces.