jh-jeong/smoothing-multiscale

Code for the paper "Multi-scale Diffusion Denoised Smoothing" (NeurIPS 2023)

27
/ 100
Experimental

This project offers a method to create more reliable image classification models by making them robust to minor input variations while maintaining high accuracy. It takes pre-trained image classifiers and training data, applies a multi-scale denoising technique, and outputs a more resilient model. This is for machine learning engineers or researchers who need to build robust vision systems.

No commits in the last 6 months.

Use this if you need to build image classification models that are highly accurate on clean data and also very robust to subtle, adversarial manipulations or noisy inputs.

Not ideal if you are looking for a plug-and-play solution for non-image data or if you lack the significant computational resources (GPUs) required for fine-tuning large models.

image-classification model-robustness computer-vision adversarial-defense deep-learning-research
Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 6 / 25
Maturity 16 / 25
Community 5 / 25

How are scores calculated?

Stars

15

Forks

1

Language

Python

License

MIT

Last pushed

Apr 30, 2024

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/jh-jeong/smoothing-multiscale"

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