RummyDetection

Introduction

RummyDetection is a Python tool which can automatically detect rummikub tiles in an image. Its purpose is to use it to automatically generate the input files for RummySolver.

It can generate both types of input files, and attempts to detect whether something looks off to make sure the user double-checks the results.

Acknowledgements

This tool draws heavily from this tutorial on detecting license plates by MicrocontrollersAndMore.

Building the tool

This project requires the Python bindings for OpenCV 3. It is highly recommended to use virtual environments for this. A very thorough guide to install OpenCV for Python in Ubuntu can be found in PyImageSearch.

Once that is set up, clone this git directory and ensure it has access to OpenCV by creating a symbolic link for the OpenCV shared object from the base OpenCV environment's site-packages into this one's. Then, install the necessary libraries from requirements.txt with pip install -r requirements.txt and you are ready to go!

Using the tool

python3 Main.py [-h] -i IMAGE -m {deck,table} [-o OUTPUT] -c CANON_DIR

Optional arguments:
  -h, --help            show this help message and exit
  -i IMAGE, --image IMAGE
                        Path to the image to be scanned
  -m {deck,table}, --mode {deck,table}
                        Whether the image represents a player's deck or the
                        table
  -o OUTPUT, --output OUTPUT
                        The output directory. If none is specified, it outputs
                        to the screen
  -c CANON_DIR, --canon_dir CANON_DIR
                        The directory containing binary images of each digit

The script requires three mandatory arguments:

• --image: the image to be scanned for tiles.
• --mode: whether the image represents a player's deck or the table. It will define the format of the output.
• --canon_dir: a directory containing a binarized version of each digit. It is used to increase the accuracy of the estimates.

Some sample images are provided in examples/.

If the user in interested in the intermediate steps leading to the final result, they can be visualized by setting show_steps to True inside Main.py.

For more information on the output format, refer to RummySolver.

About KNN

This repository contains some precomputed training data for KNN, so it can be used off-the-shelf. However, retraining is possible by using GenerateData.py, which requires a binarized image containing all characters that need to be detected. See this tutorial for more information.