Online proctoring with face analysis and object recognition using Yolo

Abstract

To guarantee academic integrity, the spread of online education and tests has prompted the creation of secure and dependable online proctoring solutions. We present a comprehensive real-time online proctoring system in this study that uses gaze movement analysis, facial aspect ratio assessment, and mouth opening status detection to detect probable cheating behaviours during remote tests. Furthermore, the system uses the You Only Look Once (YOLO) algorithm to detect forbidden goods like as phones and books within the examination area.The gaze movement analysis module monitors students' eye movements during the examination using computer vision techniques such as eye tracking algorithms built in Scipy. The technology may detect instances of prolonged off-screen gazing by analysing gaze patterns, highlighting potential attempts to access unauthorised materials.The facial aspect ratio analysis component calculates the aspect ratio of major facial features using facial landmarks. This method aids in detecting anomalous head motions or changes from conventional facial expressions that may suggest dishonesty.Furthermore, the proctoring system employs mouth opening status detection, which is accomplished through the use of deep learning algorithms, to identify instances of verbal communication or whispering throughout the exam. Our method incorporates the YOLO object detection technique, in addition to gaze and facial analysis, to recognise phones and books in the examination setting. Using a pre-trained YOLO model, the system can detect and flag any unauthorised items in the area of the students. Using a broad dataset of simulated examination scenarios and real-world proctoring instances, we assess the performance of the proposed online proctoring system. The findings illustrate the system's accuracy and effectiveness in detecting potential instances of academic cheating, giving teachers vital insights into student behaviour during remote tests. Finally, our study introduces a novel technique to real-time online proctoring that incorporates gaze movement analysis, face aspect ratio assessment, mouth opening status detection, and YOLO-based object detection. This multifaceted approach helps to improve the integrity and security of remote exams in the digital education era.