ANALYSIS OF THE IMPACT OF VIDEO QUALITY ON FEATURE EXTRACTION FROM A VIDEO STREAM USING CONVOLUTIONAL NEURAL NETWORKS

Abstract

Online proctoring is a monitoring technology that can spot any improper conduct both before and during the exam right away. Audio, video streams, and desktop screen recordings are just a few of the inputs used for surveillance. The proctor, a particularly qualified professional who checks for errors in the system and searches for infractions on the part of exam-takers, generally keeps an eye on the proctoring procedure. Nevertheless, only two persons can be under one person's supervision at once, and it is also feasible to let phony test-takers in. When creating a trainable AI network, you can use numerous variables simultaneously. We created a convolutional neural network-based deep learning approach for automatic face detection. Video samples with varied criteria yielded more accurate findings. Our model properly predicted the outcome for each sampling criterion, except for the test with accesso-ries. Other instances demonstrate 28–32 percent efficiency in determining the sample-face difference.