Skip to content

Dataset · sensenova

Multimodal image-text corpus for scene understanding

sensenova/SenseNova-Vision-Corpus-50M

Large-scale English image-text dataset for unified multimodal model training.

What's inside

The dataset was created by the authors of the Vision as Unified Multimodal Generation paper and includes English images and texts. Despite mentioning 50 million examples in the title and corpus description, the repository metadata indicates a size of less than 1,000 records. The data was collected to train models for visual understanding and geometry tasks.

What you can do with it

  • Training and fine-tuning any-to-any multimodal models
  • Improving the understanding of geometry and spatial relationships in images
  • Experiments with unified text and graphics generation
Tasks
Multimodal
Languages
English
Size
up to 1K rows
Downloads this month
9.5K
Likes
62

How to load it

from datasets import load_dataset

ds = load_dataset("sensenova/SenseNova-Vision-Corpus-50M")