mmrl/lost-in-latent-space

Code from the article: "Lost in Latent Space: Examining Failures of Disentangled Models at Combinatorial Generalisaton" (NeurIPS, 2022)

12
/ 100
Experimental

This project helps researchers and machine learning practitioners understand why generative AI models, specifically those designed to 'disentangle' different properties of an object (like its shape, color, or position), struggle to create realistic new combinations of those properties. It takes in datasets of images with varied features and outputs analyses of how well different model architectures handle creating novel combinations. This is for anyone researching or implementing advanced generative models and their ability to generalize.

No commits in the last 6 months.

Use this if you are developing or evaluating disentangled generative models and need to rigorously test their ability to produce novel combinations of features.

Not ideal if you are looking for a pre-trained model to immediately generate high-quality images or for applications not related to evaluating model generalization.

generative-ai-research machine-learning-evaluation representational-learning computer-vision-research
No License Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 4 / 25
Maturity 8 / 25
Community 0 / 25

How are scores calculated?

Stars

8

Forks

Language

Python

License

Last pushed

Jan 04, 2023

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/mmrl/lost-in-latent-space"

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