import datasets import pandas as pd _CITATION = """\ @InProceedings{huggingface:dataset, title = {anti-spoofing_Live}, author = {TrainingDataPro}, year = {2023} } """ _DESCRIPTION = """\ The dataset consists of 40,000 videos and selfies with unique people. 15,000 attack replays from 4,000 unique devices. """ _NAME = 'anti-spoofing_Live' _HOMEPAGE = f"https://huggingface.co/datasets/TrainingDataPro/{_NAME}" _LICENSE = "" _DATA = f"https://huggingface.co/datasets/TrainingDataPro/{_NAME}/resolve/main/data/" class AntiSpoofingLive(datasets.GeneratorBasedBuilder): def _info(self): return datasets.DatasetInfo( description=_DESCRIPTION, features=datasets.Features({ 'phone': datasets.Value('string'), 'selfie': datasets.Image(), 'video': datasets.Value('string'), 'worker_id': datasets.Value('string'), 'age': datasets.Value('int8'), 'country': datasets.Value('string'), 'gender': datasets.Value('string') }), supervised_keys=None, homepage=_HOMEPAGE, citation=_CITATION, ) def _split_generators(self, dl_manager): data = dl_manager.download(f"{_DATA}data.tar.gz") annotations = dl_manager.download(f"{_DATA}{_NAME}.csv") data = dl_manager.iter_archive(data) return [ datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={ "data": data, 'annotations': annotations }), ] def _generate_examples(self, data, annotations): annotations_df = pd.read_csv(annotations, sep=';') for idx, (image_path, image) in enumerate(data): if image_path.endswith('.jpg'): yield idx, { 'phone': annotations_df.loc[ annotations_df['selfie_link'] == image_path] ['phone'].values[0], 'selfie': { 'path': image_path, 'bytes': image.read() }, 'video': annotations_df.loc[ annotations_df['selfie_link'] == image_path] ['video_link'].values[0], 'worker_id': annotations_df.loc[ annotations_df['selfie_link'] == image_path] ['worker_id'].values[0], 'age': annotations_df.loc[ annotations_df['selfie_link'] == image_path] ['age'].values[0], 'country': annotations_df.loc[ annotations_df['selfie_link'] == image_path] ['country'].values[0], 'gender': annotations_df.loc[ annotations_df['selfie_link'] == image_path] ['gender'].values[0], }