A Novel CNN-RNN Model for E-Cheating Detection Based on Video Surveillance

Nowadays, everything needs to be digitized, and scientific knowledge is constantly bringing comfort and change to everyday life. Autonomous systems have become a prominent technology in recent years, with a variety of applications in different fields. Our proposed system is designed to maintain academic integrity in exams. The system uses computer vision techniques to monitor the behavior of students during the exam and detect any suspicious activities, such as looking at someone else's paper or using unauthorized materials. The proposed E-Cheating Detection consists of four core steps: 1) Student/Person Detection and Tracking, 2) Detect suspicious activities, 3) Generating alerts, and 4) Mark attendance. Student detection from videos is performed by using YOLOv7 and DeepSort tracker is used to track the detected persons that are being detected by the YOLOv7 algorithm. To classify suspicious activities (such as the exchange of paper, and giving codes to one another, etc.) the system uses CNN-RNN architecture in which the inceptionV3 model is used for feature extraction. The system will generate real-time alerts for suspicious behavior by sending an email via SMTP to the exam administration /invigilator. The system marks the student's attendance by recognizing and matching student faces that are stored in the database. The performance of the system will be evaluated by conducting a series of experiments using simulated scenarios, and the results will demonstrate the effectiveness of the proposed system in detecting suspicious activities during physical exams. The proposed system has the potential to promote exam integrity and create a fair environment for all students, ultimately improving the education system's quality. Keywords: Detect Suspicious Activities; CNN-RNN, YOLOv7; DeepSort; Face Recognition; Physical Exam; Smart Attendance