Pareidolia Engine

Application built with TypeScript, React and Node.js (Express). It detects forms in images using bounding box regions and machine generated descriptions. It can interface with remote LLM providers or vision-compatible local models.

Region detection is performed on the client side using canvas pixel manipulation, where an edge detection algorithm groups connected components into bounding box regions. This process is left deliberately without input configuration controls. From the perspective of the UI, the detection process is black-boxed. Each region is then extracted and sent to a large language model to provide descriptions.

Setup instructions can be found on the GitHub README.

Repo