import tensorflow_datasets as tfds

class UnearthedArtifacts(tfds.core.GeneratorBasedBuilder):
    VERSION = tfds.core.Version("1.0.0")

    def _info(self):
        return tfds.core.DatasetInfo(
            builder=self,
            description="Artifact image dataset.",
            features=tfds.features.FeaturesDict({
                "image": tfds.features.Image(),
                "label": tfds.features.ClassLabel(
                    names=["scrapers", "arrowheads", "axehammerstones"]
                ),
            }),
            supervised_keys=("image", "label"),
        )

    def _split_generators(self, dl_manager):
        img_dir = r"C:\Users\chava\unearthed_artifacts_dataset"
        return {
            "train": self._generate_examples(img_dir),
        }

    def _generate_examples(self, img_dir):
        import os

        classes = ["scrapers", "arrowheads", "axehammerstones"]

        for label in classes:
            class_path = os.path.join(img_dir, label)
            if not os.path.exists(class_path):
                continue

            for img_name in os.listdir(class_path):
                if img_name.lower().endswith((".jpg", ".jpeg", ".png")):
                    yield f"{label}_{img_name}", {
                        "image": os.path.join(class_path, img_name),
                        "label": label,
                    }