sail-sg/DiffMemorize

[TMLR 2025] On Memorization in Diffusion Models

32
/ 100
Emerging

This project provides tools for researchers and machine learning engineers to investigate how and why diffusion models 'memorize' training data. It allows you to input image datasets like CIFAR-10 or ImageNet, train diffusion models under various conditions, and then measure the extent to which these models reproduce specific training examples rather than generating novel images. The outputs are metrics and generated images that reveal memorization patterns.

No commits in the last 6 months.

Use this if you are a researcher or ML engineer studying the privacy implications, robustness, or generalization capabilities of diffusion models and need to empirically quantify memorization.

Not ideal if you are looking to simply train a diffusion model for image generation or apply one to a creative task, as this tool is focused on the analytical study of memorization.

diffusion-models model-auditing machine-learning-research generative-AI data-privacy
Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 7 / 25
Maturity 16 / 25
Community 9 / 25

How are scores calculated?

Stars

31

Forks

3

Language

Python

License

MIT

Last pushed

Oct 05, 2023

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/sail-sg/DiffMemorize"

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