imagewoof.py

# Copyright (C) 2022, François-Guillaume Fernandez.

# This program is licensed under the Apache License 2.0.
# See LICENSE or go to <https://www.apache.org/licenses/LICENSE-2.0> for full license details.

"""Imagewoof dataset."""

import os
import json

import datasets


_HOMEPAGE = "https://github.com/fastai/imagenette#imagewoof"

_LICENSE = "Apache License 2.0"

_CITATION = """\
@software{Howard_Imagewoof_2019,
    title={Imagewoof: a subset of 10 classes from Imagenet that aren't so easy to classify},
    author={Jeremy Howard},
    year={2019},
    month={March},
    publisher = {GitHub},
    url = {https://github.com/fastai/imagenette#imagewoof}
}
"""

_DESCRIPTION = """\
Imagewoof is a subset of 10 classes from Imagenet that aren't so 
easy to classify, since they're all dog breeds. The breeds are: 
Australian terrier, Border terrier, Samoyed, Beagle, Shih-Tzu, 
English foxhound, Rhodesian ridgeback, Dingo, Golden retriever, 
Old English sheepdog.
"""

_LABEL_MAP = [
    'n02086240',
    'n02087394',
    'n02088364',
    'n02089973',
    'n02093754',
    'n02096294',
    'n02099601',
    'n02105641',
    'n02111889',
    'n02115641',
]

_REPO = "https://huggingface.co/datasets/frgfm/imagewoof/resolve/main/metadata"


class ImagewoofConfig(datasets.BuilderConfig):
    """BuilderConfig for Imagewoof."""

    def __init__(self, data_url, metadata_urls, **kwargs):
        """BuilderConfig for Imagewoof.
        Args:
          data_url: `string`, url to download the zip file from.
          matadata_urls: dictionary with keys 'train' and 'validation' containing the archive metadata URLs
          **kwargs: keyword arguments forwarded to super.
        """
        super(ImagewoofConfig, self).__init__(version=datasets.Version("1.0.0"), **kwargs)
        self.data_url = data_url
        self.metadata_urls = metadata_urls


class Imagewoof(datasets.GeneratorBasedBuilder):
    """Imagewoof dataset."""

    BUILDER_CONFIGS = [
        ImagewoofConfig(
            name="full_size",
            description="All images are in their original size.",
            data_url="https://s3.amazonaws.com/fast-ai-imageclas/imagewoof2.tgz",
            metadata_urls={
                "train": f"{_REPO}/imagewoof2/train.txt",
                "validation": f"{_REPO}/imagewoof2/val.txt",
            },
        ),
        ImagewoofConfig(
            name="320px",
            description="All images were resized on their shortest side to 320 pixels.",
            data_url="https://s3.amazonaws.com/fast-ai-imageclas/imagewoof2-320.tgz",
            metadata_urls={
                "train": f"{_REPO}/imagewoof2-320/train.txt",
                "validation": f"{_REPO}/imagewoof2-320/val.txt",
            },
        ),
        ImagewoofConfig(
            name="160px",
            description="All images were resized on their shortest side to 160 pixels.",
            data_url="https://s3.amazonaws.com/fast-ai-imageclas/imagewoof2-160.tgz",
            metadata_urls={
                "train": f"{_REPO}/imagewoof2-160/train.txt",
                "validation": f"{_REPO}/imagewoof2-160/val.txt",
            },
        ),
    ]

    def _info(self):
        return datasets.DatasetInfo(
            description=_DESCRIPTION + self.config.description,
            features=datasets.Features(
                {
                    "image": datasets.Image(),
                    "label": datasets.ClassLabel(
                        names=[
                            "Australian terrier",
                            "Border terrier",
                            "Samoyed",
                            "Beagle",
                            "Shih-Tzu",
                            "English foxhound",
                            "Rhodesian ridgeback",
                            "Dingo",
                            "Golden retriever",
                            "Old English sheepdog",
                        ]
                    ),
                }
            ),
            supervised_keys=None,
            homepage=_HOMEPAGE,
            license=_LICENSE,
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):
        archive_path = dl_manager.download(self.config.data_url)
        metadata_paths = dl_manager.download(self.config.metadata_urls)
        archive_iter = dl_manager.iter_archive(archive_path)
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    "images": archive_iter,
                    "metadata_path": metadata_paths["train"],
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split.VALIDATION,
                gen_kwargs={
                    "images": archive_iter,
                    "metadata_path": metadata_paths["validation"],
                },
            ),
        ]

    def _generate_examples(self, images, metadata_path):
        with open(metadata_path, encoding="utf-8") as f:
            files_to_keep = set(f.read().split("\n"))
        idx = 0
        for file_path, file_obj in images:
            if file_path in files_to_keep:
                label = _LABEL_MAP.index(file_path.split("/")[-2])
                yield idx, {
                    "image": {"path": file_path, "bytes": file_obj.read()},
                    "label": label,
                }
                idx += 1