nis-research/nn-frequency-shortcuts

Frequency Shortcuts in Neural Networks

26
/ 100
Experimental

This project helps computer vision researchers and practitioners understand how neural networks classify images, specifically by analyzing which image frequencies (like textures or shapes) they prioritize. It takes in trained image classification models and image datasets, then outputs metrics and visualizations that reveal whether the network relies on 'frequency shortcuts.' This insight helps those trying to build more robust and generalizable image classification systems.

No commits in the last 6 months.

Use this if you are a computer vision researcher or ML engineer investigating why your image classification models sometimes fail on new, unseen data and suspect they might be learning superficial patterns.

Not ideal if you are looking for a plug-and-play solution to directly improve model performance without needing to understand the underlying learning mechanisms.

image-classification model-robustness computer-vision-research neural-network-analysis deep-learning-diagnostics
Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 6 / 25
Maturity 16 / 25
Community 4 / 25

How are scores calculated?

Stars

21

Forks

1

Language

Python

License

MIT

Last pushed

Nov 01, 2024

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/nis-research/nn-frequency-shortcuts"

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