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 [optional]

@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}
}

We add Arabic to the dataset and rename it as XTD11. The effort of the Arabic dataset goes to:

@misc{aggarwal2020zeroshot,
      title={WASF: An open-source dataset for Image-Text},
      author={Khalid Almubarak, Mohammed Alkhowaiter, Asem  Alasaker},
      year={2020},
      eprint={2012.05107},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}