Cheat Checker

Promoting fair knowledge practices with the help of computer vision.

Comment

Inspiration 🌟

Our inspiration sprang from our own struggles - a few of us had been unjustly accused of cheating by our professors, and others had worked tirelessly while some of our peers simply took shortcuts and got credit for our efforts. We wanted to make a project to address that issue and make a base for a computer vision-powered program that helps professors make fair judgments.

*We believe that pursuing knowledge is a valuable and important aspect of education, and it should be carried out with integrity and honesty. Cheating not only undermines the learning process but also diminishes the value of hard work and effort put in by those who strive for academic excellence. We wanted to use new AI capabilities to help the teachers and students themselves. *

What it does ⚙️

Cheat Checker is a web-based application that uses computer vision to process the submitted videos from the student during the test-taking. Python model is trained to detect when the student is not looking at the screen and when their gaze is absent. Our machine learning model detects when the student is not looking at the screen, is absent, and puts out the timestamps for that activity. Teachers can see for themselves the video and check with the provided data to make a fair judgment.

How we built it 🐍

Django framework, HTML/CSS, Python(2 and 3) libraries, Pytorch.

Challenges we ran into 🤔

  • Learning and implementing the Django framework from scratch was a real challenge. Connecting front-end and back-end with Django. Debugging issues with styles.css and other web development things.
  • Different versions of Python interpreters. Set up of Pytorch.
  • Defining the "cheating" for the training model. And merging two machine learning models, one with depth estimation and the other with vertical/horizontal ratio, into one three-dimensional eye-tracking model.
  • We don't have LiDAR, depth laser measurement, to enhance the accuracy of the model that will make the model usable for different accessibility cases.
  • Designing the front-end.

Accomplishments that we're proud of 🔥

  • We developed a working code for uploading the videos from the user into the server as well as connecting the HTML/CSS to the backend using Django.
  • We extended the functionality of the existing pre-trained eye tracking mode: 1) by adding a dynamic real-time eye detection ratio 2) model can be tested not only on web camera but on videos as well

What we learned ✅

  • We learned how to implement the Django framework and connect the back-end and front-end.
  • Knowledge about different versions of Python interpreters.
  • How to modify existing AI tools.
  • How to work as a team and have fun in the process.

What's next for Cheat Checker 🌠

We acknowledge users with disabilities, which is why we want to continue modifying the model by adding a depth calculator into it to make the accuracy better and be able to give different exceptions. We are concerned about users' privacy and think this is an important issue, so we wanted to look into how to enhance students' privacy as well. Overall, we want to continue working on all aspects of our project as it has grown close to our hearts.

Built With

Share this project:

Updates