This is a script to extract important information from pictures of Maimai results screens, for use in whatever kind of tracker you want to create.
To get all the Python dependencies for this project, simply run
pip install -r requirements.txt
This project requires Tesseract https://github.com/tesseract-ocr/tesseract to be installed on the machine. If the tesseract installation is not automatically found by the script, you can add a config.ini file with contents:
[DEFAULT]
tesseract_path = <full/path/to/executable/tesseract>
Which will show the script where to find it. Your Tesseract installation will need both the eng.traineddata and jpn.traineddata files (eng should be installed by default). TODO: explain how to get these on different systems
This project requires OpenCV to be installed on the machine. TODO: point to OpenCV installation details
The main function for processing is processImg. It takes in an image as an array (e.g. from cv2.imread) and returns a python dictionary with the extracted data. It also takes an optional argument (imshow) with a list of strings for displaying some partially processed images for debugging purposes. The valid imshow strings are:
'ref_points': Show the detected reference points and the screen bounding circle. These are used to determine the cropping for detecting other features'num_edges': Show the edge only version of the score, used for matching'num_detect': Show the detected numbers and their locations. Number value corresponds with resistor color codes'title_processed': Show the fully processed versions of the title. This is what is fed into OCR'diff_crop': Show the small cropped region used for reading the difficulty level
So an example call might be:
#process the image and display the reference points and processed titles
result = processImg(image,imshow=['ref_points','title_processed'])Results screen images to be processed by this should contain the entire screen (the circle of the screen is used for a refrence point), and they should be fairly head-on (no perspective correction is done). TODO: Add example images
The returned result contains the following fields:
'score': The score, as an integer. Your actual percent score isscore/100.'titles': A list of predicted titles, from most to least likely. Each title is given as a tuple of(name, id, confidence).'difficulty': The difficulty, as a string. Will always be one of the standard 6 difficulties.
There are some utility functions in testMaimai.py that are useful for local testing and debugging.
testFiletakes a file name in as a string + an optional imshow argument. It processes the file and prints the results.testAlltakes a directory name in as a string + an optional imshow argument. It processes all of the files in that folder, printing the file name and results for each of them.