emnist_mnist.py

import struct

import numpy as np

import datasets
from datasets.tasks import ImageClassification


_URL = "./raw/"
_URLS = {
    "train_images": "emnist-mnist-train-images-idx3-ubyte.gz",
    "train_labels": "emnist-mnist-train-labels-idx1-ubyte.gz",
    "test_images": "emnist-mnist-test-images-idx3-ubyte.gz",
    "test_labels": "emnist-mnist-test-labels-idx1-ubyte.gz",
}


class EMNIST(datasets.GeneratorBasedBuilder):

    BUILDER_CONFIGS = [
        datasets.BuilderConfig(
            name="emnist-mnist",
            version=datasets.Version("1.0.0"),
        )
    ]

    def _info(self):
        return datasets.DatasetInfo(
            features=datasets.Features(
                {
                    "image": datasets.Image(),
                    "label": datasets.features.ClassLabel(
                        names=["0", "1", "2", "3", "4", "5", "6", "7", "8", "9"]
                    ),
                }
            ),
            supervised_keys=("image", "label"),
            task_templates=[
                ImageClassification(
                    image_column="image",
                    label_column="label",
                )
            ],
        )

    def _split_generators(self, dl_manager):
        urls_to_download = {key: _URL + fname for key, fname in _URLS.items()}
        downloaded_files = dl_manager.download_and_extract(urls_to_download)
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    "filepath": (
                        downloaded_files["train_images"],
                        downloaded_files["train_labels"],
                    ),
                    "split": "train",
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split.TEST,
                gen_kwargs={
                    "filepath": (
                        downloaded_files["test_images"],
                        downloaded_files["test_labels"],
                    ),
                    "split": "test",
                },
            ),
        ]

    def _generate_examples(self, filepath, split):
        """This function returns the examples in the raw form."""
        # Images
        with open(filepath[0], "rb") as f:
            # First 16 bytes contain some metadata
            _ = f.read(4)
            size = struct.unpack(">I", f.read(4))[0]
            _ = f.read(8)
            images = np.frombuffer(f.read(), dtype=np.uint8).reshape(size, 28, 28)

        # Labels
        with open(filepath[1], "rb") as f:
            # First 8 bytes contain some metadata
            _ = f.read(8)
            labels = np.frombuffer(f.read(), dtype=np.uint8)

        for idx in range(size):
            yield idx, {"image": images[idx], "label": str(labels[idx])}