utm_dataset.py

import os
from datasets import DatasetInfo, Features, ClassLabel, Image, GeneratorBasedBuilder, Split

class UTM_Dataset(GeneratorBasedBuilder):
    """UTM Dataset organized in train/validation/test splits with subfolders as classes."""

    VERSION = "1.0.0"

    def _info(self):
        """Returns the dataset metadata, features, and supervised keys."""
        return DatasetInfo(
            description="UTM Dataset organized in train/validation/test with subfolders as classes",
            features=Features(
                {
                    "image": Image(),  # images
                    "label": ClassLabel(names=self._get_class_names())  # labels from subfolder names
                }
            ),
            supervised_keys=("image", "label"),
        )

    def _get_class_names(self):
        """Get class names from the train folder subdirectories."""
        train_dir = os.path.join(self.config.data_dir, "train")
        return sorted(
            [d for d in os.listdir(train_dir) if os.path.isdir(os.path.join(train_dir, d))]
        )

    def _split_generators(self, dl_manager):
        """Defines the splits and their corresponding folders."""
        data_dir = self.config.data_dir

        return [
            Split.TRAIN: self._generate_examples(os.path.join(data_dir, "train")),
            Split.VALIDATION: self._generate_examples(os.path.join(data_dir, "validation")),
            Split.TEST: self._generate_examples(os.path.join(data_dir, "test")),
        ]

    def _generate_examples(self, path):
        """Yields (id, example) tuples for each image in the folder."""
        for class_name in sorted(os.listdir(path)):
            class_dir = os.path.join(path, class_name)
            if not os.path.isdir(class_dir):
                continue
            for fname in sorted(os.listdir(class_dir)):
                if fname.lower().endswith((".png", ".jpg", ".jpeg")):
                    # Unique id for each example
                    yield f"{class_name}_{fname}", {
                        "image": os.path.join(class_dir, fname),
                        "label": class_name,
                    }