SATIN.py

import datasets
from datasets import load_dataset


_CONSTITUENT_DATASETS = ['SAT-4', 'SAT-6', 'NASC-TG2', 'WHU-RS19', 'RSSCN7', 'RS_C11', 'SIRI-WHU', 'EuroSAT',
                         'NWPU-RESISC45', 'PatternNet', 'RSD46-WHU', 'GID', 'CLRS', 'Optimal-31',
                         'Airbus-Wind-Turbines-Patches', 'USTC_SmokeRS', 'Canadian_Cropland',
                         'Ships-In-Satellite-Imagery', 'Satellite-Images-of-Hurricane-Damage',
                         'Brazilian_Coffee_Scenes', 'Brazilian_Cerrado-Savanna_Scenes', 'Million-AID',
                         'UC_Merced_LandUse_MultiLabel', 'MLRSNet',
                         'MultiScene', 'RSI-CB256', 'AID_MultiLabel']


class SATINConfig(datasets.BuilderConfig):
    """BuilderConfig for SATIN"""

    def __init__(self, name, **kwargs):

        super(SATINConfig, self).__init__(version=datasets.Version("1.0.0"), **kwargs)
        self.name = name
        self.hf_dataset_name = 'jonathan-roberts1' + "/" + name
        self.description = None
        self.features = None


class SATIN(datasets.GeneratorBasedBuilder):
    """SATIN Images dataset"""

    BUILDER_CONFIGS = [SATINConfig(name=dataset_name) for dataset_name in _CONSTITUENT_DATASETS]

    def _info(self):
        if self.config.description is None or self.config.features is None:
            stream_dataset_info = load_dataset(self.config.hf_dataset_name, streaming=True, split='train').info
            self.config.description = stream_dataset_info.description
            self.config.features = stream_dataset_info.features
        return datasets.DatasetInfo(
            description=self.config.description,
            features=self.config.features,
        )

    def _split_generators(self, dl_manager):
        dataset = load_dataset(self.config.hf_dataset_name)
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={"data_path": dataset},
            ),
        ]

    def _generate_examples(self, data_path):
        # iterate over the Huggingface dataset and yield the idx, image and label
        _DEFAULT_SPLIT = 'train'
        huggingface_dataset = data_path['train']
        features = huggingface_dataset.features
        for idx, row in enumerate(huggingface_dataset):
            features_dict = {feature: row[feature] for feature in features}
            # Reorder features to make image the first feature
            image = features_dict.pop('image')
            features_dict = {'image': image, **features_dict}
            yield idx, features_dict