giannisdaras/ambient-diffusion

[NeurIPS 2023] Official Implementation: "Ambient Diffusion: Learning Clean Distributions from Corrupted Data"

34
/ 100
Emerging

Ambient Diffusion helps researchers and scientists working with image data that is inherently damaged or incomplete. It takes highly corrupted image datasets and learns the underlying clean image distribution. The output is a generative model capable of creating new, uncorrupted images that reflect the true data, even if the original clean data was never observed. This tool is ideal for scientists, medical imaging specialists, or anyone dealing with low-quality or incomplete visual information.

No commits in the last 6 months.

Use this if you need to understand and generate clean images from datasets where every single sample is significantly corrupted, such as images with missing pixels or blockages.

Not ideal if your dataset is mostly clean with only minor, occasional corruptions, or if you primarily need to clean individual images rather than learn an entire data distribution.

medical-imaging scientific-imaging data-generation image-reconstruction corrupted-data-analysis
No License Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 9 / 25
Maturity 8 / 25
Community 17 / 25

How are scores calculated?

Stars

85

Forks

16

Language

Python

License

Last pushed

Oct 23, 2023

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/giannisdaras/ambient-diffusion"

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