FakeOrNot

Inspiration

Deepfake models have advanced significantly in recent years. We wanted to create a way for people to test their ability to spot deepfakes and see how they measure up against state-of-the-art models.

What it does

We developed a game where users select up to five words. Based on these words, the user is shown either an AI-generated or a real image. The task is to guess whether the image is real or fake. Afterward, users can see if their guess was correct and compare it to the detection model’s prediction.

How we built it

The frontend was built using Vue, while the backend is powered by Django. We use OpenAI’s API for image generation and SerpAPI to retrieve real images. For detection, we integrated the SightEngine API.

Challenges we ran into

One of the main challenges was finding an API that retrieves real images from the web rather than entire web pages.

Accomplishments that we're proud of

We developed a solid plan early in the hackathon and stuck to it. We also successfully set up the web app within a short timeframe while collaborating efficiently as a team.

What we learned

We learned the importance of prioritizing key tasks and leaving additional features for after the minimum viable product was complete. We also gained experience in implementing and working with various APIs.