Skip to content

Dataset · pufanyi

Multilingual image and text dataset

pufanyi/MIMICIT

Millions of image and text examples across six languages for training multimodal models.

What's inside

It contains images and texts in English, Chinese, Spanish, Japanese, French, and Korean. The corpus size ranges from 1 to 10 million records. The set was prepared by researchers from Nanyang Technological University (S-Lab) and Microsoft Research, with data converted from JSON to parquet format for fast loading.

What you can do with it

  • Training and SFT fine-tuning of multilingual vision-language models (VLMs)
  • Training cross-lingual image-text retrieval and matching systems
Languages
English, Chinese, Spanish, Japanese, French, Korean, Arabic
Size
1M–10M rows
Downloads this month
857
Likes
49
Access
on request on Hugging Face

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("pufanyi/MIMICIT")