Skip to content

Dataset · OpenFace-CQUPT

10 million human image captions

OpenFace-CQUPT/HumanCaption-10M

Large-scale English image-text pair dataset for generating and recognizing images of people.

What's inside

The dataset includes 10 to 100 million pairs of human images and their natural English text descriptions. Built on top of Laion-face data, the original images can be accessed after completing a usage agreement on GitHub and following the instructions on Hugging Face.

What you can do with it

  • Training text-to-image models for more accurate generation of people from text descriptions
  • Training image-to-text systems and multimodal models for detailed description of appearance and actions
Tasks
Image → text, Text → image
Languages
English
Size
10M–100M rows
Downloads this month
190
Likes
48

How to load it

from datasets import load_dataset

ds = load_dataset("OpenFace-CQUPT/HumanCaption-10M")