Automatic Attendance Camera for Classrooms

^{I worked on this project under Professor Mithuna Thottethodi from Fall 2017 to Spring 2018 and this acted as my senior design project. The idea was to create an automatic system to take attendance using any camera and facial recognition. There was an app component to the project but I mainly worked on the facial recognition portion.}

^{For the facial recognition portion we would find regions of interest for the faces in the image and then obtain the bounding boxes for those RoIs. After that we would pass in the face in that bounding box to a pretrained network like FaceNet or OpenFace. These would output an array that was the encoded information from the face and then this was used in conjunction with another classifer to do the facial recognition to determine who was in class.}

^{Since one image could potentially give false information, we utilized multiple images. The use of multiple images allowed us to determine if a student was detected in some given percentage of images that we could confidently say they were present in class.}

^{Another thing we took advantage of when detecting whether a student was present was the idea of locality in classes. Students tend not to move once they sit down in their class for the day, so we could use the idea of locality to get rid of some misclassified faces in a few frames.}

^{For more information on this here is the project page}