NovaGuard

Generated Heat Map (backend)
Metrics (landing page)
Landing page
Another Heat Map (backend)

Inspiration

One very common and concurrent issue which we foresee only worsening that plagues people when viewing art or general entertainment on line is deciding whether the art made was genuine or whether it was spawned via AI. It is this detection problem that we sought to solve.

What it does

Our code takes short form content as an input from the user, once in the back we "grade" it against metrics in which AI tends to perform poorly as well as generate heat maps to circle the regions that point to it being created by AI. Once finished, we compute a weighed sum of those grades, and output these grades to the user as well as the confidentially score that determines whether it was created by AI or not.

How we built it

First we researched what goes into a video being "AI" like in nature and then decided upon computationally how do we then measure these metrics as well as how do we give them an arbitrary score. We decided on clarity, texture balance, stability, compression grid, and color consistency. After which we wanted to build a dynamic and sleek front end that would explain to the user their score, what the metrics measure as well a final diagnostic report that gives it the results of their input.

Challenges we ran into

Some challenges that we ran into were deployment, API development, creating and training our Machine Learning model, and finding the limitations of our project.

Accomplishments that we're proud of

We were proud of our project's accuracy in grading videos as it was able to achieve 70% accuracy against a mix of 20 real and blatantly AI generated videos in less than 36 hours. Moreover we are really proud of our website design and the beauty with which our heat maps came out with on the backend.

What we learned

We learned how to coordinate well as a team with git as well as using more professional techniques such as pushing, pulling, committing, and the like to keep a clean record of the project's history. We learned better about how to actually deploy a website.

What's next for NovaGuard

Our main goal is to pursue entrepreneurship and develop some new piece of technology involving AI that seeks to revolutionize what people think the limits of AI are. In the meantime, we seek to continue building by attending more hackathons, particularly research-based hackathons so our coding prowess may increase.