I built an AI vision workspace for poker table recognition
LumiGap is a macOS app for building poker table recognition workflows.
It reads visible table data from the screen, uses OCR and Core ML to recognize things like names, stacks, bets, cards, board, pot, and table regions, then turns that into structured data you can review, correct, export, and use for custom datasets.
The interesting part is the pipeline;
recognize table state, correct it, build datasets, train or connect models, and test them on real layouts.
It is built for poker researchers, coaches, ML experimenters, and advanced players who want their own AI-powered research workflow.
It is not a poker room, betting product, or real-time play assistant; users are responsible for following the rules of any platforms they use.
