Inspiration
We believe that there isn't enough material online on helping people get better at spotting AI images. While there are tools to check if it's AI, you probably won't use those while scrolling social media. Therefore, we decided that a better way was needed to help train people on how to determine if an image is generated by AI or not.
What it does
It provides a user with different courses, which are collections of images based on categories that are often faked or edited with AI (such as vehicles, people, purchasable goods, etc.). Each user has a name, password, email, and elo. For every correctly distinguished image, a user can gain or lose elo. Images themselves also have elo, which increases the more often users incorrectly guess them and decreases the more often users correctly guess them. The amount of elo a user receives from correctly distinguishing an image depends on the image elo; the higher the image elo (The tougher that image is to distinguish), the more elo the user gains.
Elo is then used as a means of ranking people on the leaderboard compared to other active users.
*Users can be created, signed in to, altered, and deleted.
How we built it
Frontend is built with React and Tailwind CSS, backend is built with Java Springboot.
For the ai images used, we generated them using Google's gemini, adobe firefly, and thispersondoesnotexist.com.
For the non-ai images, we used various sources from the internet, all credited and linked to in the sources.txt file in the assets folder.
Information for users, courses, and images is stored on an SQL database. We followed OAuth 2.0 Password Grant authentication to ensure that user data remains private and secure throughout each session.
Challenges we ran into
Establishing the exact look and feel of the website Getting the database to initially connect to the website The authentication was tough to implement at times
Accomplishments that we're proud of
We're proud of the overall look and feel of the website, including storing user information, utilizing authentication to provide secure connections, creating an elo rating system, and ultimately delivering a functioning end product that helps address an issue we believe is prevalent in today's society.
What we learned
How to create proper user Authentication, make a functioning elo system, and small improvements on existing knowledge in the tools we used.
What's next for AI Spotter
We'd add lessons (giving people information on the different categories and common tells in each of them), and add sections for videos generated by AI and text generated by AI. (we would also add more categories, and more images)
Log in or sign up for Devpost to join the conversation.