Inspiration
Pokemon, AR games, collecting stuff.
What it does
A mobile web application that you can use to monsterify ordinary objects with your camera, and collect a menagerie of Scribbles.
How we built it
4 separate servers, running frontend, image processing and classification, data management and routing, and a database. It uses local image segmentation algorithm to remove the background and extract the monster, and then classifies the monsters with the OpenAI GPT-4 API.
The frontend was written in NextJS, and the backend servers were flask, with Redis for storage.
Challenges we ran into
OpenAI credit usage. Collaborative coding on a very fast moving project making git difficult. With the added struggles of last minute integration between all the parts.
Accomplishments that we're proud of
Image segmentation and decoration works really well. UI is pretty and effective.
The deployment server pulls the git repo automatically and hot reloads on changes, which means we are always using the latest version available in the deployment, this helped speed up the rapid prototyping!
What we learned
To keep project scope manageable. And that react and canvas elements don't play well.
What's next for Scribble Scraps
Adding battling between Scribbles. We almost have it working.
Built With
- javascript
- nextjs
- openai
- python
- react
- redis
Log in or sign up for Devpost to join the conversation.