deepaction_v1.py

import os
import datasets

_CITATION = """\
@inproceedings{tbd,
  title={tbd},
  author={tbd},
  year={2024},
  url={https://huggingface.co/datasets/faridlab/deepaction_v1}
}
"""

_DESCRIPTION = """\
TBD
"""

_HOMEPAGE = "https://huggingface.co/datasets/faridlab/deepaction_v1"

_LICENSE = "TBD"

SUPPORTED = ["VideoPoet", "BDAnimateDiffLightning", "CogVideoX5B", "Pexels", "RunwayML", "StableDiffusion", "Veo"]

class DeepActionV1(datasets.GeneratorBasedBuilder):
    VERSION = datasets.Version("1.0.0")

    def _info(self):
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=datasets.Features({
                "video": datasets.Video(),
                "label": datasets.ClassLabel(names=SUPPORTED),  # Add all category names
            }),
            supervised_keys=("video", "label"),
            homepage=_HOMEPAGE,
            license=_LICENSE,
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):
        urls = []
        base_url = "https://huggingface.co/datasets/faridlab/deepaction_v1/resolve/main/"

        for engine in SUPPORTED:
            for i in list(range(95)) + list(range(99, 104)):
                if engine in ["Pexels", "Veo"]:
                    urls.append(os.path.join(base_url, engine, "{}".format(i), "a.mp4"))
                elif engine in ["VideoPoet"]:
                    for vid_file in ["a.mp4", "b.mp4", "c.mp4", "d.mp4"]:
                        urls.append(os.path.join(base_url, engine, "{}".format(i), vid_file))
                else:
                    for vid_file in ["a.mp4", "b.mp4", "c.mp4", "d.mp4", "e.mp4"]:
                        urls.append(os.path.join(base_url, engine, "{}".format(i), vid_file))
        
        data_dir = dl_manager.download_and_extract(urls)

        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={"data_dir": data_dir},
            ),
        ]

    def _generate_examples(self, data_dir):
        for label in os.listdir(data_dir):
            label_path = os.path.join(data_dir, label)
            if os.path.isdir(label_path):
                for subfolder in os.listdir(label_path):
                    subfolder_path = os.path.join(label_path, subfolder)
                    if os.path.isdir(subfolder_path):
                        for video_file in os.listdir(subfolder_path):
                            if video_file.endswith(".mp4"):
                                video_path = os.path.join(subfolder_path, video_file)
                                yield video_path, {
                                    "video": video_path,
                                    "label": label,
                                }