Haon-Chen/mmE5-synthetic

mmE5 Synthetic Data

This dataset contains synthetic datasets used for the finetuning of mmE5 (mmE5: Improving Multimodal Multilingual Embeddings via High-quality Synthetic Data):

• Classification
• Retrieval
• VQA

Github

Image Preparation

First, you should prepare the images used for training:

Image Downloads

• Download Links: Download image resources for dataset via the following link:
  • LAOIN-images

Image Organization

  images/
  ├── laion/
        └── ... .jpg
  

You can refer to the image paths in each subset to view the image organization.

You can also customize your image paths by altering the image_path fields.

Citation

If you use this dataset in your research, please cite the associated paper.

@article{chen2025mmE5,
  title={mmE5: Improving Multimodal Multilingual Embeddings via High-quality Synthetic Data},
  author={Chen, Haonan and Wang, Liang and Yang, Nan and Zhu, Yutao and Zhao, Ziliang and Wei, Furu and Dou, Zhicheng},
  journal={arXiv preprint arXiv:2502.08468},
  year={2025}
}