ygo_card_img.py

import datasets
import json 
import os 

_CITATION = ''

_DESCRIPTION = ''

_URL = 'https://huggingface.co/datasets/akumoth/ygo_card_img/'

_IMAGES_DIR = 'imgs/'

class ImageSet(datasets.GeneratorBasedBuilder):

    IMAGE_EXTENSION = ".png"

    def _info(self):
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=datasets.Features(
                {
                    "id": datasets.Value("int32"),
                    'name': datasets.Value('string'),
                    'japanese_name': datasets.Value('string'),
                    'img': datasets.Image(),
                    'card_type': datasets.Value('string'),
                    'attribute': datasets.Value('string'),
                    'archetypes': datasets.Value('string'),
                }
            ),
            supervised_keys=None,
            homepage='https://huggingface.co/datasets/akumoth/ygo_card_img',
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):
        downloaded_files = dl_manager.download({
            'card_json': 'card_df.jsonl',
            'images': 'card_imgs.zip'
        })
        return [ datasets.SplitGenerator(
                    name=datasets.Split.TRAIN,
                    gen_kwargs={
                        'card_json': downloaded_files['card_json'],
                        'image_iter': dl_manager.iter_archive(downloaded_files['images']),
                    }
                ),
                datasets.SplitGenerator(
                    name=datasets.Split.VALIDATION,
                    gen_kwargs={
                        'card_json': downloaded_files['card_json'],
                        'image_iter': dl_manager.iter_archive(downloaded_files['images']),
                    }
                ),
                ]

    def _generate_examples(self, card_json, image_iter):
        examples = [json.loads(example_json) for example_json in open(card_json).readlines()]
        for img_path, img_obj in image_iter:
            if img_path.startswith(_IMAGES_DIR):
                cid = int(''.join(filter(str.isdigit, img_path)))
                examples[cid]['img'] = img_obj.read()
                id_ = examples[cid]['id']
                yield id_, examples[cid]