Inspiration
Over the years, I have seen a variety of methods used by faculty to take attendance in class and during exams. These include physical sign-in sheets, clicker systems and/or require T.A.s to physically check student IDs. These methods prove to be cumbersome and lack robustness.
What it does
The system performs facial recognition to recognise students in a class and sends confirmation emails with a picture of the recognised person to the desired faculty member in real-time. The students need to pass by the device whilst looking into the camera for a brief period (2s) which allows quick attendance taking. The integrity of the attendance is also ensured as false sign-ins are extremely difficult due to the robustness of the algorithm and datasets.
How we built it
Built around a raspberry pi 3, the system uses a simple USB webcam and a facial recognition neural network to cross reference people in the video stream with a database of student pictures.
Accomplishments that we're proud of
The device accomplishes embedded machine learning efficiently and is able to recognise up to 10 students simultaneously by utilising a cascade filters, deep learning, and transfer learning methods to speed up training and performance.
What's next for A.I. Assisted Attendance and Academic Integrity System
Upgrades to the current raspberry pi used will improve recognition speeds tenfold and a camera setup capable of autofocus will allow attendance taking for seated students in large auditoriums without the need for stepping directly in front of the camera.
Log in or sign up for Devpost to join the conversation.