j-min/DallEval

DALL-Eval: Probing the Reasoning Skills and Social Biases of Text-to-Image Generation Models (ICCV 2023)

36
/ 100
Emerging

Provides comprehensive evaluation frameworks across four dimensions: visual reasoning skills via DETR-based object detection and spatial understanding, demographic biases (gender/skin tone) in generated outputs, image quality through FID scoring, and image-text alignment using CLIP retrieval and VL-T5 captioning. Includes inference implementations for multiple text-to-image models (DALL-E variants, minDALL-E, Stable Diffusion) enabling cross-model benchmark comparisons.

143 stars. No commits in the last 6 months.

Stale 6m No Package No Dependents
Maintenance 2 / 25
Adoption 10 / 25
Maturity 16 / 25
Community 8 / 25

How are scores calculated?

Stars

143

Forks

6

Language

Jupyter Notebook

License

MIT

Last pushed

Jun 10, 2025

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/j-min/DallEval"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.