Novosibirsk State Technical University (NETI) has proposed its own version of the "ballot box" for universities. The program, developed by students of the Faculty of Automation and Computer Engineering, is able to recognize faces. This will simplify the voting procedure and speed up the counting of votes.
A team of students from the Faculty of Automation and Computer Engineering at NSTU-NETI has developed an automated voting system in universities using computer vision technology. The device is powered by Raspberry Pi, has a camera and a display, and allows you to vote only after recognizing a person's face and confirming the identity of the person who wants to leave their vote.
According to Nikita Mishechkin, a third—year student at AUTF, the development combines applied programming, computer vision and an urgent social task - to ensure transparency and ease of voting. "The camera captures the user's face, and then computer vision algorithms based on OpenCV and dlib/face_recognition read the face and translate it into a unique numeric vector (embedding), which is then compared with other embeddings in the database of faces. If the user has already voted, re—authorization is not possible," Nikita Mishechkin said.
The step-by-step voting procedure looks like this: the user selects the desired vote and the answer option, the smart "ballot box" recognizes the face, the system checks in the database whether this user has already voted. If you did not vote, your vote is counted. All votes are automatically counted, and the results are displayed as charts and graphs for the administrator without the usual paper ballots and tedious manual counting. The whole procedure takes 10-15 seconds.
The "ballot box" using computer vision technology is designed for student elections. According to Nikita Mishechkin, voting at the university is an optimal test site, but it does not exclude the possibility of using the development outside the university. "Modern technologies create an engagement effect — it's interesting. But the main factor is trust. It is impossible to vote twice even for very similar participants: we use numerical vectors of facial features, not just visual similarity," Nikita Mishechkin emphasized.
The automated voting system is almost ready, and the team needs to conduct operational tests. Possible technical risks are reduced due to proven libraries, unit tests and stress testing of the system, and there are no risks of data leakage: after scanning, faces are stored as embeddings, not images, and encrypted, while the database is protected from changes.
Earlier, students of the Faculty of Automation and Computer Engineering of NSTU-NETI presented a website with a map of hiking routes in Novosibirsk that will be of interest to both citizens and visitors of Novosibirsk, as well as a navigator with artificial intelligence for recognizing obstacles that will help blind and visually impaired people navigate the terrain.