yasserben/DATUM

[CVPR-W 2023] Official Implementation of One-shot Unsupervised Domain Adaptation with Personalized Diffusion Models

33
/ 100
Emerging

When you have a segmentation model trained on one type of visual data (like synthetic images) and need it to work well on a new, unlabeled real-world dataset (like street scenes), this project helps. It takes a single example image from your target domain and uses advanced AI to generate many new, photo-realistic training images that perfectly match the style and diverse contexts of that new domain. This allows an AI researcher or computer vision engineer to adapt their existing segmentation models to perform accurately on new, challenging visual data.

No commits in the last 6 months.

Use this if you need to rapidly adapt an image segmentation model to a new visual domain with very little data, achieving strong performance without extensive manual labeling.

Not ideal if you have abundant labeled data for your target domain or if your task does not involve adapting computer vision models across different visual styles or environments.

computer-vision image-segmentation domain-adaptation synthetic-data-generation machine-learning-research
Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 9 / 25
Maturity 16 / 25
Community 8 / 25

How are scores calculated?

Stars

78

Forks

4

Language

Python

License

Last pushed

Jan 09, 2024

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/yasserben/DATUM"

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