Netrunner Card Recognition
Run the script with python makerseye_v0.2.py.
Requires a bunch of packages (cv2, numpy, pickle, PIL, urllib, imagehash), most which can be installed with pip install [package]. Google for more informaton.
scans.32ihash is the datafile and should be in the same directory as the script.
Requires an internet connection to jinteki.net to retrieve detected images.
When the script starts running, it will open a number of windows:
img: The raw webcam input
do: What the script sees.
can: Output of the edge detection to detect the card.
crop: Output of what the script thinks is the card.
match: Displays the matched card. First opens when a card is detected.
You can click and drag in the img window to define a zone of interest, which the script will only look within. The zone of interest is bounded by a red box.
While the script is running, there are a number of global keyboard commands that you can use on any active window:
ESC: Ends the script.
b: Zeroes the current image, treating the current background as baseline. Use this if your blank background shows any white in the can window.
c: Clears the background defined by b, as well as any defined zone of interest.
h: Hides/Shows the red zone of interest box in the img window.
i: Invert the image. Use only if are running into detection issues.
Lines 16-18 in the script have a number of parameters that you can edit.
LANG: Supports 'en' and 'zh-simp', defines which language of image is pulled from jinteki.net.
STDEVS: Defines how strict the recognition algorithm is before it returns a match. The recognition result must be at least STDEVS standard deviations better than the average recognition result. Higher is more strict, leading to more accuracy in recognition, but also more failures to recognize.
CROP_PX: Number of pixels to crops from the borders of the card before attempting recognition. Sometimes helps if the script consistently detects large sleeves.