查看文章

Online examinations have turned out to be the new normal. However, it is not that easy to proctor the students as rigorously as in in-center examinations. It is essential to find an approach to proctor the online examinations too as rigorously as possible. There are already several webcam proctoring systems that are used in the real world, but these systems are not very accurate and miss out on detecting all possible malpractices and in certain cases due to defect in the system it detects a malpractice for someone who never even attempted any. This project focuses mainly on building features that can make the existing webcam proctoring system more advanced and rigorous. The project is aimed at building the following features namely head pose estimation, mouth opening detection, eye ball monitoring, number of persons detection, mobile phone detection and face spoofing detection. For each of these features, machine learning models are built using Python. All these features make use of the live webcam feed which is obtained using OpenCV and an output is obtained which gives information about the direction of the head and eyes, presence of more than one person and presence of mobile phone, opening of mouth, occurrence of face spoofing. All these outputs are recorded as a log file which can be used to identify any possible malpractices based on these features.