xd-violence-small.py

import os
import datasets
import numpy as np

class YourDataset(datasets.GeneratorBasedBuilder):
    VERSION = datasets.Version("1.0.0")

    BUILDER_CONFIGS = [
        datasets.BuilderConfig(
            name="i3d_rgb",
            version=VERSION,
            description="i3d RGB dataset",
            features=datasets.Features({"video": datasets.Sequence(datasets.Value("float32"))}),
        ),
        datasets.BuilderConfig(
            name="i3d_flow",
            version=VERSION,
            description="i3d Flow dataset",
            features=datasets.Features({"video": datasets.Sequence(datasets.Value("float32"))}),
        ),
        datasets.BuilderConfig(
            name="vggish_train",
            version=VERSION,
            description="VGGish Train dataset",
            features=datasets.Features({"audio": datasets.Sequence(datasets.Value("float32"))}),
        ),
        datasets.BuilderConfig(
            name="vggish_test",
            version=VERSION,
            description="VGGish Test dataset",
            features=datasets.Features({"audio": datasets.Sequence(datasets.Value("float32"))}),
        ),
    ]

    def _info(self):
        features = self.config.features
        return datasets.DatasetInfo(
            features=features,
        )

    def _split_generators(self, dl_manager):
        data_dir = "Alkemi/xd-violence-small"

        splits = []
        for config in self.configs:
            if config.name.startswith("i3d-features"):
                split_name = config.name.split("_")[1]
                splits.append(
                    datasets.SplitGenerator(name=split_name, gen_kwargs={"data_dir": os.path.join(data_dir, "i3d-features", split_name)})
                )
            elif config.name.startswith("vggish-features"):
                split_name = config.name.split("_")[1]
                splits.append(
                    datasets.SplitGenerator(name=split_name, gen_kwargs={"data_dir": os.path.join(data_dir, "vggish-features", split_name)})
                )

        return splits

    def _generate_examples(self, data_dir):
        filepaths = []

        for root, dirs, files in os.walk(data_dir):
            for file in files:
                if file.endswith(".npy"):
                    filepaths.append(os.path.join(root, file))

        for i, filepath in enumerate(filepaths):
            data = np.load(filepath)

            yield i, {"video": data}  # or {"audio": data} depending on the configuration