renovation.py

import os

import datasets
from datasets.tasks import ImageClassification

_HOMEPAGE = "https://github.com/your-github/renovation"

_CITATION = """\
@ONLINE {renovationdata,
    author="Your Name",
    title="Renovation dataset",
    month="January",
    year="2023",
    url="https://github.com/your-github/renovation"
}
"""

_DESCRIPTION = """\
Renovations is a dataset of images of houses taken in the field using smartphone
cameras. It consists of 3 classes: cheap, average, and expensive renovations.
Data was collected by the your research lab.
"""

_URLS = {
    "cheap": "https://huggingface.co/datasets/rshrott/renovation/resolve/main/cheap.7z",
    "average": "https://huggingface.co/datasets/rshrott/renovation/resolve/main/average.7z",
    "expensive": "https://huggingface.co/datasets/rshrott/renovation/resolve/main/expensive.7z",
}

_NAMES = ["cheap", "average", "expensive"]


class Renovations(datasets.GeneratorBasedBuilder):
    """Renovations house images dataset."""

    def _info(self):
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=datasets.Features(
                {
                    "image_file_path": datasets.Value("string"),
                    "image": datasets.Image(),
                    "labels": datasets.features.ClassLabel(names=_NAMES),
                }
            ),
            supervised_keys=("image", "labels"),
            homepage=_HOMEPAGE,
            citation=_CITATION,
            task_templates=[ImageClassification(image_column="image", label_column="labels")],
        )

    def _split_generators(self, dl_manager):
        data_files = dl_manager.download_and_extract(_URLS)
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    "files": dl_manager.iter_files([data_files["cheap"]]),
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split.VALIDATION,
                gen_kwargs={
                    "files": dl_manager.iter_files([data_files["average"]]),
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split.TEST,
                gen_kwargs={
                    "files": dl_manager.iter_files([data_files["expensive"]]),
                },
            ),
        ]

    def _generate_examples(self, files):
        for i, path in enumerate(files):
            file_name = os.path.basename(path)
            if file_name.endswith(".jpg"):
                yield i, {
                    "image_file_path": path,
                    "image": path,
                    "labels": os.path.basename(os.path.dirname(path)).lower(),
                }