Dataset · dinhanhx
Multilingual image description dataset
dinhanhx/crossmodal-3600
Set of 3,600 global images with reference descriptions in 36 languages.
What's inside
It is a diverse set of 3,600 images from around the world, equipped with human-generated descriptions in 36 languages (including Arabic, Bengali, Danish, German, and others). Created for evaluating the quality of multilingual image descriptions.
What you can do with it
- Benchmarking multilingual image-to-text models
- Evaluating image caption generation quality
- Tasks
- Image → text
- Languages
- Arabic, Bangla, Czech, Danish, German, Greek, English, Spanish and 4 more
- Size
- 1K–10K rows
- Downloads this month
- 1.4K
- Likes
- 7
How to load it
from datasets import load_dataset
ds = load_dataset("dinhanhx/crossmodal-3600")