test_images_demo.py

import datasets

_CITATION = """\
@InProceedings{huggingface:dataset,
title = {Small htr examples images},
author={Gabriel Borg},
year={2023}
}
"""

_DESCRIPTION = """\
Demo dataset for the htr demo.
"""
_HOMEPAGE = "https://huggingface.co/datasets/Riksarkivet/test_images_demo"

_LICENSE = ""

_REPO = "https://huggingface.co/datasets/Riksarkivet/test_images_demo"

class ExampleImages(datasets.GeneratorBasedBuilder):
    """Small sample of image-text pairs"""

    def _info(self):
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=datasets.Features(
                {
                    'text': datasets.Value("string"),
                    'image': datasets.Image(),
                }
            ),
            supervised_keys=None,
            homepage=_HOMEPAGE,
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):
        images_archive = dl_manager.download(f"{_REPO}/resolve/main/images.tar.gz")
        metadata_path = dl_manager.download(f"{_REPO}/resolve/main/images.txt")
        image_iters = dl_manager.iter_archive(images_archive)
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    "images": image_iters,
                    "metadata_path": metadata_path
                }
            ),
        ]

    def _generate_examples(self, images, metadata_path):
        """Generate images and text."""
        with open(metadata_path, encoding="utf-8") as f:
            metadata_list = f.read().split("\n")
        for idx, (img_obj, meta_txt) in enumerate(zip(images, metadata_list)):
            filepath, image = img_obj

            text_value = meta_txt.split("= ")[-1].strip()

            yield idx, {
                "image": {"path": filepath, "bytes": image.read()},
                "text": text_value,
            }