Skip to content

Dataset · common-canvas

High-res images with synthetic descriptions

common-canvas/commoncatalog-cc-by-sa

Millions of photos up to 4K with Creative Commons licenses and synthetic English captions.

What's inside

A large multimodal corpus ranging from 1 million to 10 million records, collected from 2014 Yahoo Flickr user photographs under various Creative Commons licenses. The key feature of the set is the high resolution of images up to 4K. Synthetic English text descriptions are added to the images.

What you can do with it

  • Training high-resolution text-to-image generation models
  • Training LoRA adapters and visual encoders on open data without strict licensing limitations
Tasks
Text → image
Languages
English
Size
1M–10M rows
Downloads this month
32.7K
Likes
12

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("common-canvas/commoncatalog-cc-by-sa")