magic_card_detector

Python implementation of a Magic: the Gathering card segmentation and recognition

General description of the algorithms and several example can be found in my blog: https://tmikonen.github.io/quantitatively/2020-01-01-magic-card-detector/

TODO:

• Expects a folder and not an image name to analize. it will not run any detection if you give it a jpg file name because it appends it to the path you supply.
• Generate phash for other sets
• Remove hard-coded paths
• actually do arg pasring for reference images
• put results.txt into the output folder.
• figure out what is different between my reference images and the ones used. resolution?
• remove path reuiqrement from MagicCardDetector. It is not used by save_hash.py.

Usage

1. Download all images and lists of cards with scryfalldler

ls sets | parallel php scryfalldler -s {} -debug
mv *.csv sets_list

2. save hash of dataset To generate the phash of a set of cards, for example March of the Machines with set code MOM and all images in a folder of the same name, run the following:

python save_hash.py ../scryfalldler/sets/MOM/ MOM.dat

You can speed this up with gnu parallel:

ls ../scryfalldler/sets | parallel "python save_hash.py ../scryfalldler/sets/{}/ dat/{}.dat"

3. Scan cards into folders labled by set To scan images in batches from an Auto Document Feeder scanner. Ensure that the images are zero padded so that the scripts can read them in the proper order.

scanimage -d epsonds:libusb:002:004 --format=jpeg --batch=card_%04d.jpg -x 70 -y 90 --batch-start=1

4. identify cards To identify cards in ../mycards and output results to ../mycards_identified, run the following:

python magic_card_detector.py --phash MOM.dat ../mycards/MOM/ ../mycards_identified/MOM/

To run this in parallel with gnu parallel:

ls ../mycards/ | parallel python magic_card_detector.py {} ../mycards/{}/ ../mycards_identified/{}/ dat/ &> log.txt

5. map identified cards to collector number

ls ../mycards_identified/ | parallel python map_cards.py {} ../scryfalldler/sets_list/ ../mycards_identified/ ../collection

Scryfalldler

Use scryfalldler to get the images for each of the sets. I added a patch to also print a mapping from image name to collector number. This will get exported to a csv which we will use later to map into the final csv for import into card catalogue of choice.

While you can run scryfalldler with parallel, I would advise against it. The API documentaion for scryfalldler says they will ban IPs who hammer the server too hard and we don't want to annoy them for providing so much to the community.

To only print the csv information, use -debug to skip the download part. You can run this in parallel, because you arent actually downloading anything here; only printing out information.

ls sets | parallel php scryfalldler -s {} -debug