This is our group project for face recognition based attendence system. We implemented this system first using OpenCV only doing everything and then using dlib for face detection and OpenCV for face recognition but we did not get much accuracy in both. Link to the OpenCV only project is here and link to our second attempt with OpenCV and dlib is here.
We then tried using OpenFace but since neural networks require high graphic support and we were trying to do this project without any server support for later general use, so we did not went that way.
We then used dlib for face detection and microsoft's vision cognitive api for face recognition and got a accuracy of 99% in a classroom of 55 students. We have a video testing this system in a classroom of 27 students uploaded here : https://youtu.be/FeNasBaXdhg This accuracy is because of the microsoft's neural network formed on our dataset. We used the camera of the phone OenPlus 3T.
More updates later.