Aim: 

Achieving exam integrity through an AI-driven Smart Proctoring System for vigilant monitoring and prevention of malpractices in online assessments.

Abstract:

From past three years all human activities are affected by COVID19 but where the knowledge gaining and education for students aren’t stopped. Due to covid-19 affect many of the educational institutions are providing online degrees and most of the students are actively taking part in advanced online courses. Conducting online examination and proctoring all the students to stop malpractice by a single proctor is a major challenging task for educational institutions. Day by day as the technology is enlarging, with the advanced AI technology we can proctor all the students with less human effort. In this project we lively monitor the activity of test taker like face verification, eye blink, head pose estimation and eye movements. This AI integrated system will prevent the cheating in examination without physical presence of the human proctor.

Proposed System:

       In light of the overfitting challenges associated with CNNs, the proposed system introduces a novel approach to Smart Proctoring. Leveraging advanced deep learning models, particularly pre-trained models such as FaceNet512, the proposed system aims to enhance the accuracy and robustness of online exam surveillance. FaceNet512, renowned for its effectiveness in face verification tasks, provides a foundation for reliable identity verification during exams. Complementing this, Support Vector Regression (SVR) is employed for precise head pose estimation, contributing to a comprehensive understanding of the test-taker's engagement. Furthermore, head poses, eye-related behaviors, including blinks and movements, are adeptly identified using the Mediapipe library. By integrating these sophisticated components, the proposed system seeks to overcome the overfitting concerns of the existing CNN-based approach, ensuring a more adaptable, accurate, and inclusive Smart Proctoring experience.

Advantages:

        FaceNet512, chosen for face verification, offers a distinctive advantage in its ability to generate highly discriminative facial embeddings. Its deep architecture excels in capturing intricate facial features, ensuring robust and accurate identity verification during online exams. FaceNet512's versatility and efficiency make it a standout choice for maintaining exam integrity.

      Support Vector Regression (SVR) contributes a significant advantage through its unparalleled precision in head pose estimation. SVR excels in accurately capturing the subtle nuances of a test-taker's head movements, providing a nuanced understanding of engagement. This precision enhances the overall reliability of the Smart Proctoring System, particularly in scenarios where nuanced head poses are indicative of user attention.

     Mediapipe stands out for its versatility in eye blink and movement identification. Its real-time, multi-modal capabilities offer a comprehensive solution for tracking eye behaviors during online exams. The efficiency of Mediapipe in handling head pose direction, dynamic eye movements, blinks, and gaze direction sets it apart, providing a robust means of assessing user engagement and potential irregularities.