import requests from pathlib import Path from typing import List import datasets logger = datasets.logging.get_logger(__name__) _URLS = {'train': 'https://huggingface.co/datasets/nateraw/auto-cats-and-dogs/resolve/main/train.tar.gz'} _NAMES = ['cat', 'dog'] class ImageFolder(datasets.GeneratorBasedBuilder): def _info(self): return datasets.DatasetInfo( features=datasets.Features( dict( file=datasets.Value("string"), labels=datasets.features.ClassLabel(names=_NAMES) ) ), ) def _split_generators(self, dl_manager: datasets.DownloadManager) -> List[datasets.SplitGenerator]: data_files = dl_manager.download_and_extract(_URLS) if isinstance(data_files, str): return [datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs=dict(archive_path=data_files))] splits = [] for split_name, folder in data_files.items(): splits.append(datasets.SplitGenerator(name=split_name, gen_kwargs=dict(archive_path=folder))) return splits def _generate_examples(self, archive_path): labels = self.info.features['labels'] extensions = set(('.jpg', '.jpeg', '.png', '.ppm', '.bmp', '.pgm', '.tif', '.tiff', '.webp')) for i, path in enumerate(Path(archive_path).glob('**/*')): if path.suffix in extensions: yield i, dict(file=path.as_posix(), labels=labels.encode_example(path.parent.name.lower()))