A ridiculously simple app where AI judges your hand-drawn doodles. It's a fun experiment to showcase the powers of Cartesi Coprocessor with Eigenlayer infrastructure. We use an inference model trained over the popular QuickDraw dataset to judge doodles.
On a high level, a user submits a doodle and a theme to our on-chain smart contract - ScribbleTaskManager. The role of this task manager is to relay the tasks to our Cartesi Coprocessor that's running our core application logic. The backend runs the doodle over a pre-trained AI model and outputs the top predictions back to the smart contract.
To run the app, you'll need to spin up the backend and then run the UI.
cartesi-backend: Navigate to this directory and follow the instructions in the README. You'll have Coprocessor Devnet environment running in no time with our Python app code and corresponding Smart contracts deployed.
ui: Navigate to this directory and follow the instructions in the README.