Combining Deep Residual Neural Network Features With Supervised Machine Learning Algorithms For Real-time Face Recognition-based Intelligent Systems

Face Recognition is the domain of technology in Computer Vision that deals with the process of identifying faces of known and unknown persons-based on facial patterns. Despite all the recent researches these years on Face Recognition technology, the development of real-time face recognition has always been a challenging task. This kind of technology has its applications widespread like security, medical diagnosis, educational sectors, etc. The advancements in High-end processors and High-Definition cameras led to the design of Face recognition systems that use offline or real-time input datasets. In this paper, our main aim is to focus on real-time video feeds taken from a framed classroom environment to identify the students or the faculty members and tag their names, com- paring them with the already stored face databases. Attendance marking is a daily routine that follows calling the name or passing the attendance books, which is very timeconsuming, and they tend to begin proxy at-tendances. This study proposes a attendance marking system using face recognition and Deep Learning techniques on a Raspberry Pi board. The proposed system delivers an approach to make real-time face recognition-based attendance systems by extracting deep facial features using deep residual network (ResNet)-based CNNs. Then that deep facial feature dataset is combined with Machine Learning Algorithm such as SVM to perform face detection and recognize the faces. The maximum recognition accuracy of 80% is obtained using the planned system on the real- time face images provided and will be further improved.