Handwritten_Khatt.py

import datasets
import pandas as pd
import os
import logging

_DESCRIPTION = """\
Arabic Handwritten dataset.
"""

_REPO = "https://huggingface.co/datasets/eDaraty/Handwritten_Khatt"


# logging.basicConfig(level=logging.INFO)
# logger = logging.getLogger(__name__)


class Khatt(datasets.GeneratorBasedBuilder):
    """Handwritten arabic image-text pairs"""

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

    def _split_generators(self, dl_manager):
        train_archive = dl_manager.download(f"{_REPO}/resolve/main/data/train.zip")
        test_archive = dl_manager.download(f"{_REPO}/resolve/main/data/test.zip")
        val_archive = dl_manager.download(f"{_REPO}/resolve/main/data/validation.zip")
        # split_metadata_paths = dl_manager.download(_METADATA_URLS)

        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    "images": dl_manager.iter_archive(train_archive)
                },
            ),

            datasets.SplitGenerator(
                name=datasets.Split.TEST,
                gen_kwargs={
                    "images": dl_manager.iter_archive(test_archive)
                },
            ),

            datasets.SplitGenerator(
                name=datasets.Split.VALIDATION,
                gen_kwargs={
                    "images": dl_manager.iter_archive(val_archive)
                },
            ),
        ]

    def _generate_examples(self, images):
        """ This function returns the examples in the raw (text) form."""
        df = pd.read_csv(f"{_REPO}/resolve/main/data/metadata.csv")

        for idx, (filepath, image) in enumerate(images):
            image_name = os.path.basename(filepath)
            # logger.info(" filename of image is '%s' ", image_name)

            description = df[df["file_name"] == image_name]['text'].values.tolist()[0]
            # logger.info(" text of image is '%s' ", description)
            # logger.info(" type of image is '%s' ", type(description))
            yield idx, {
                "image": {"path": filepath, "bytes": image.read()},
                "text": description,
            }