Wan-fen52/MAVOS-DD

LICENSE: This dataset is released under the CC BY-NC-SA 4.0 license.

This repository contains MAVOS-DD an open-set benchmark for multilingual audio-video deepfake detection.

Below, you can find the code to obtain the subsets described in the paper: train, validation, open-set model, open-set language and open-set full:

from datasets import Dataset, concatenate_datasets
metadata = Dataset.load_from_disk('MAVOS-DD')
metadata_indomain = metadata.filter(lambda sample: sample['split']=='test' and not sample['open_set_model'] and not sample['open_set_language'])
metadata_open_model = metadata.filter(lambda sample: sample['split']=='test' and sample['open_set_model'] and not sample['open_set_language'])
metadata_open_model = concatenate_datasets([metadata_indomain, metadata_open_model])
metadata_open_language = metadata.filter(lambda sample: sample['split']=='test' and not sample['open_set_model'] and sample['open_set_language'])
metadata_open_model = concatenate_datasets([metadata_indomain, metadata_open_language])
metadata_all = metadata.filter(lambda sample: sample['split']=='test')

The scripts require the datasets package to be installed.

pip install datasets

We provide two scripts: metadata_generation.py and dataset.py. The metadata_generation.py script is responsible for generating the metadata. Below is a sample metadata entry:

Sample: {'video_path': 'arabic/inswapper/02690.png_Po82BhllEjA_340_1.mp4.mp4', 'label': 'fake', 'split': 'train', 'open_set_model': False, 'open_set_language': False, 'language': 'arabic', 'generative_method': 'inswapper'}

The dataset.py script includes examples of how to read and filter this metadata.

The code for running the baseline models can be found here: https://github.com/CroitoruAlin/MAVOS-DD

Note: Our dataset was collected from publicly available YouTube videos. If any individual wishes to request the removal of content involving them, please contact us at alincroitoru97@gmail.com.

Citation:

@misc{Croitoru-ArXiv-2025,
      title={MAVOS-DD: Multilingual Audio-Video Open-Set Deepfake Detection Benchmark}, 
      author={Florinel-Alin Croitoru and Vlad Hondru and Marius Popescu and Radu Tudor Ionescu and Fahad Shahbaz Khan and Mubarak Shah},
      year={2025},
      eprint={2505.11109},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2505.11109}, 
}