kidney-ct-abnormality.py

import datasets
import numpy as np
import pandas as pd
from pathlib import Path
import os
import json

_CITATION = """\
Alves, N., & Boulogne, L. (2023). Kidney CT Abnormality [Data set]. Zenodo. https://doi.org/10.5281/zenodo.8043408
"""

_DESCRIPTION = """\
This dataset in total is comprised of 986 .mha (medical high-resolution image) files. Each of these files contains multiple layers of CT scans of the kidney. 
The dataset has been speparated into train and test set by initial processing.

"""

_HOMEPAGE = "https://zenodo.org/records/8043408"

_LICENSE = "CC BY-NC-SA 4.0"

_URL = "https://huggingface.co/datasets/Euniceyeee/kidney-ct-abnormality"

_URLS = {
    'train': "https://huggingface.co/datasets/Euniceyeee/kidney-ct-abnormality/resolve/main/train.zip",
    'test': "https://huggingface.co/datasets/Euniceyeee/kidney-ct-abnormality/resolve/main/test.zip"
}

_METADATA_URL = {
    "metadata": "https://huggingface.co/datasets/Euniceyeee/kidney-ct-abnormality/resolve/main/dataset_m.json"
}

LABELS = [
    'Normal',
    'Abnormal'
]


class KidneyCTAbnormality(datasets.GeneratorBasedBuilder):
    """Collection of brain xray images for fine-grain classification."""
    VERSION = datasets.Version("1.0.0")

    def _info(self):
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=datasets.Features(
                {
                    "image": datasets.Image(),
                    "label": datasets.ClassLabel(num_classes=2, names=LABELS),
                }
            ),
            supervised_keys=("image", "label"),
            homepage=_HOMEPAGE,
            citation=_CITATION,
            license=_LICENSE
        )

    def _split_generators(self, dl_manager):
        metadata_url = dl_manager.download(_METADATA_URL)
        files_metadata = {}
        with open(metadata_url["metadata"], encoding="utf-8") as f:
            for lines in f.read().splitlines():
                file_json_metdata = json.loads(lines)
                files_metadata.setdefault(file_json_metdata["split"], []).append(file_json_metdata)
        
        img_url = _URLS
        data_dir = dl_manager.download_and_extract(img_url)

        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    "filepath": data_dir['train'],
                    "metadata": files_metadata["train"]
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split.TEST,
                gen_kwargs={
                    "filepath": data_dir['train'],
                    "metadata": files_metadata["test"]
                },
            ),
        ]

    def _generate_examples(self, download_path, metadata):
        """Generate images and labels for splits."""
        for i, meta in enumerate(metadata):
            img_path = os.path.join(
                download_path,
                meta["split"],
                meta["abnormality"],
                meta["file_name"],
            )
            yield i, {
                "image": img_path,
                "label": meta["abnormality"],
            }