metadata

task_categories:
  - image-to-text
  - text-to-image
language:
  - ar
  - en
  - ru
  - it
  - es
  - ko
  - pl
  - tr
  - zh
  - de
  - fr
  - ja
size_categories:
  - 1K<n<10K

Dataset Summary

The expanded XTD-11 dataset, now including Arabic, enhances the original XTD collection. This dataset introduces a 1,000-image multi-lingual MSCOCO2014 caption to test multimodel in zeroshot image or text retrieval in 11 Languages.

Dataset Details

Citation

@misc{aggarwal2020zeroshot,
      title={Towards Zero-shot Cross-lingual Image Retrieval},
      author={Pranav Aggarwal and Ajinkya Kale},
      year={2020},
      eprint={2012.05107},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}

For the Arabic version you can cite this repisotry :

@inproceedings{al2024araclip,
  title={AraCLIP: Cross-Lingual Learning for Effective Arabic Image Retrieval},
  author={Al-Barham, Muhammad and Afyouni, Imad and Almubarak, Khalid and Elnagar, Ashraf and Turky, Ayad and Hashem, Ibrahim},
  booktitle={Proceedings of The Second Arabic Natural Language Processing Conference},
  pages={102--110},
  year={2024}
}


@misc{XTD_11,
  title = XTD_11,
  author={Khalid Almubarak, Mohammed Alkhowaiter, Asem Alaskar},
  howpublished = {\url{https://huggingface.co/datasets/khalidalt/xtd_11}},
  note = {Accessed: 2023-02-02}
}