Trustworthy-ML-Lab/posthoc-generative-cbm

[CVPR 2025] Concept Bottleneck Autoencoder (CB-AE) -- efficiently transform any pretrained (black-box) image generative model into an interpretable generative concept bottleneck model (CBM) with minimal concept supervision, while preserving image quality

33
/ 100
Emerging

This tool helps researchers and designers understand and control what features appear in images generated by AI models. You input a pre-trained image generative model and it produces an interpretable version where you can see and manipulate high-level concepts (like 'smiling' or 'wearing glasses') without needing to retrain the original model or label many images. This is useful for anyone working with AI art or synthetic data who needs to ensure generated images meet specific conceptual criteria.

Use this if you need to gain interpretable control over a pre-existing image generation AI, allowing you to easily adjust specific features or characteristics in the output images.

Not ideal if you are looking to build a generative model from scratch or if you do not need to understand and control the conceptual underpinnings of your image generation process.

AI-art synthetic-data image-generation model-interpretability computer-vision-research
No License No Package No Dependents
Maintenance 10 / 25
Adoption 6 / 25
Maturity 8 / 25
Community 9 / 25

How are scores calculated?

Stars

17

Forks

2

Language

Jupyter Notebook

License

Last pushed

Mar 03, 2026

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/generative-ai/Trustworthy-ML-Lab/posthoc-generative-cbm"

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