yk7333/d3po

[CVPR 2024] Code for the paper "Using Human Feedback to Fine-tune Diffusion Models without Any Reward Model"

39
/ 100
Emerging

This project helps AI researchers and practitioners refine their image generation models to better align with specific human preferences or aesthetic goals. You input an existing diffusion model and human feedback (either quantifiable metrics or direct human annotations on generated images), and it outputs a fine-tuned diffusion model that produces images more aligned with those preferences. This is ideal for those working on improving the subjective quality or specific characteristics of AI-generated imagery.

244 stars. No commits in the last 6 months.

Use this if you need to fine-tune an image diffusion model based on human aesthetic judgments or specific quality criteria without the complexity of training a separate reward model.

Not ideal if you are looking for a tool to generate images from scratch without any fine-tuning, or if your primary goal is not related to improving image quality through feedback.

AI image generation creative AI image quality enhancement human-in-the-loop AI generative AI research
Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 10 / 25
Maturity 16 / 25
Community 13 / 25

How are scores calculated?

Stars

244

Forks

18

Language

Python

License

MIT

Last pushed

Apr 06, 2024

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/yk7333/d3po"

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