This project presents a Computer Vision (CV) based solution designed to enhance cheating detection during exams in classroom settings at Arizona State University (ASU). By leveraging CV technology to monitor students' head and eye movements, the solution aims to ensure academic integrity, promote honesty, and encourage the development of fair study habits.
Maintaining academic integrity during exams is a persistent challenge in educational institutions. Traditional surveillance methods may fail to detect covert cheating behaviors, such as secret interactions between students or sneaking glances at unauthorized materials. This project seeks to address these gaps using computer vision technology.
Implementing a real-time monitoring system using CV can help in:
- Enhanced cheating detection: Monitoring student behavior more effectively.
- Fostering a culture of honesty: Building trust between students and faculty.
- Promoting integrity: Encouraging students to focus on fair study habits.
- Stakeholders: Faculty members, Office of Academic Integrity, Information Technology Department.
- Beneficiaries: Students and the academic community benefiting from fair assessments.
Current solutions like browser lockdowns and seating arrangements cannot detect subtle behavioral indicators of cheating, such as:
- Sneaking glances at unauthorized materials.
- Secret interactions between students.
- Lack of live feedback for proctors during the exam.
Our proposed solution uses CV algorithms to track and analyze students' head and eye movements in real-time during exams. This system identifies suspicious behavior such as:
- Continuously looking away from the front view.
- Glancing in directions unrelated to the exam task.
The system employs high-resolution webcams to capture data, which is processed in real-time to detect and flag potential cheating patterns.
- Real-Time Monitoring: Captures students' head and eye movements.
- Immediate Analysis: CV model analyzes live feed for cheating indicators.
- Live Alerts: Sends alerts when potential cheating is detected.
- Intervention Protocol: Proctors are alerted to take real-time action.
- Confirmation & Action: Proctors confirm and act on the identified cheating cases.
- YOLOv4 - Object Detection
- Detects human figures and objects such as phones or notes in real time.
- Tracks the presence and location of unauthorized materials.
- Shape_Predictor_68_Face_Landmarks - Face Detection
- Monitors eye movement direction and head orientation to detect cheating behavior.
- Narrow Object Detection: Limited object recognition capabilities.
- Limited Accuracy: Potential operational errors in identifying cheating behavior.
- Device Usage Limitation: Limitations on compatible devices and hardware.
- Facial landmarks with dlib: https://pyimagesearch.com
- YOLO - Object Detection: https://pjreddie.com/darknet/yolo/
- OpenCV for face detection: OpenCV Documentation