Dataset · ScienceOne-AI
Scientific multimodal image-text pairs
ScienceOne-AI/S1-MMAlign
24.17M image-text pairs from 3.01M open-access English scientific papers.
What's inside
The dataset contains over 24.17 million high-quality image-text pairs gathered from approximately 3.01 million open-access scientific publications. The material covers various scientific disciplines and is intended for multimodal training. The language is English.
What you can do with it
- Training multimodal models for scientific images
- Evaluating VQA and image-to-text on scientific data
- Feature extraction for scientific illustration search
- Tasks
- Image → text, Visual question answering, Embeddings
- Languages
- English
- Size
- 10M–100M rows
- Downloads this month
- 6.9K
- Likes
- 107
How to load it
from datasets import load_dataset
ds = load_dataset("ScienceOne-AI/S1-MMAlign")