AI-Based Proctored Exam Tool

Inspiration

With the rise of virtual hackathons and coding competitions, the need for a fair and secure online screening process became evident. Many organizers struggle to ensure test integrity when shortlisting candidates remotely. We were inspired to create a tool that uses AI to detect cheating behaviors in real time—bringing transparency and trust to remote evaluations.

What it does

The AI-Based Proctored Exam Tool allows organizers to conduct secure online exams by:

Detecting emotions (e.g., stress, fear) using face-api.js
Identifying unauthorized objects (e.g., mobile phones) using coco-ssd
Tracking body posture and movements via PoseNet
Running in-browser with live webcam feed—no external software required It flags suspicious behavior, helping organizers shortlist genuine and focused candidates for national and international hackathons.

How we built it

We developed the tool using:

React.js for building the frontend interface
TensorFlow.js, face-api.js, and coco-ssd for AI-based detection
PoseNet for real-time pose estimation
HTML5 video and canvas APIs for webcam input and visual overlays
Modular components for scalability and easy integration into existing platforms

Challenges we ran into

Managing real-time performance without lag while processing AI models in the browser
Ensuring accuracy in emotion and object detection with limited training data
Dealing with browser compatibility for webcam and canvas rendering
Balancing user privacy with effective proctoring

Accomplishments that we're proud of

Successfully integrated multiple AI models to run simultaneously in-browser
Built a fully functional prototype with real-time detection and visual feedback
Created a tool that is both lightweight and scalable for real-world use
Developed a clean and intuitive interface for both candidates and examiners