Skip to content

Dataset · Rapidata

Human quality evaluation of generated images

Rapidata/text-2-image-Rich-Human-Feedback

Over 1.5 million ratings from real people for images generated from English text prompts.

What's inside

Contains pairs of images and English text descriptions (totaling between 10,000 and 100,000 entries). Over 152,000 people evaluated the generated images for style, consistency, and prompt alignment, while also noting defects. Data collected via the Rapidata platform.

What you can do with it

  • Training reward models for RLHF during fine-tuning of image generators
  • Training models for automated quality assessment and defect detection on images
  • Comparing and filtering results of text-to-image models
Tasks
Text → image, Text classification, Image classification, Image → text, Image segmentation
Languages
English
Size
10K–100K rows
Downloads this month
787
Likes
39

How to load it

from datasets import load_dataset

ds = load_dataset("Rapidata/text-2-image-Rich-Human-Feedback")