/deepPoker

iOS app that uses computer vision model trained using Turi Create for real-time card detection and provides probability of poker hands

Primary LanguageSwift

DeepPoker

DeepPoker is an app that supports people playing Poker for the first time by determing their strongest hand.
The app levergaes a computer vision model so that the user does not have to type into the app the card they have.

This repository has been made to help anyone who is trying to build a CoreML model for an iOS app.
In this project I have labelled a set of card images using RectLabel on a mac and have provided a script in xmlTOcsv.py to parse Rectlabel's output to the expected TuriCreate format

Getting started

If you have not already labelled your image to build a model, this is the first thing I would do. (Or use a labelled public set if this is just for experimentation)

Then convert them to the csv format that TuriCreate is expecting.
You can use combineCsv.py if you have a collection of CSV annotaitions.

Finally in prep.py change IMAGES_DIR and csv_path to meet your data setup.
And then run prep.py

prep.py will export a .sframe file.
In train.py change the directory of the data variable to point to your .sframe file.
Finally, run train.py

Demo of the DeepPoker App

License

Under MIT's License of free usage and distribution