Skip to content

Dataset · moonworks

Artistic image generation benchmark

moonworks/lunara-art-eval

8,000 generated images with prompts and evaluations from eight text-to-image models to compare their artistic qualities.

What's inside

The dataset contains 8,000 images, original English text prompts, generator model names, and three ratings for each image. The sample presents results from eight models, including GPT-Image-1-mini, HiDream-I1, Qwen-Image, and Flux-Klein. The data evaluates artistic quality, emotional expressiveness, and content integrity.

What you can do with it

  • Evaluation and comparison of image generators by artistic criteria
  • Training reward models and aesthetic image classifiers
  • Analysis of strengths and weaknesses of modern text-to-image architectures
Tasks
Text → image
Size
1K–10K rows
Downloads this month
937
Likes
45

How to load it

from datasets import load_dataset

ds = load_dataset("moonworks/lunara-art-eval")