PeterWang512/AttributeByUnlearning
Code for the paper "Data Attribution for Text-to-Image Models by Unlearning Synthesized Images."
This project helps researchers and practitioners understand which specific training images were most critical in generating a particular image from a text-to-image model. It takes a synthesized image and the trained model, then identifies the original training data that most influenced its creation. This is useful for anyone needing to trace the origins or assess the impact of specific data on AI-generated visuals.
No commits in the last 6 months.
Use this if you need to pinpoint the exact training images responsible for a specific output generated by a text-to-image AI model, without needing to retrain the model from scratch.
Not ideal if your goal is to simply evaluate the overall performance of a text-to-image model or to attribute outputs to broad categories of data rather than individual images.
Stars
17
Forks
1
Language
Python
License
—
Category
Last pushed
May 23, 2025
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/PeterWang512/AttributeByUnlearning"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
PRIS-CV/DemoFusion
Let us democratise high-resolution generation! (CVPR 2024)
mit-han-lab/distrifuser
[CVPR 2024 Highlight] DistriFusion: Distributed Parallel Inference for High-Resolution Diffusion Models
Tencent-Hunyuan/HunyuanPortrait
[CVPR-2025] The official code of HunyuanPortrait: Implicit Condition Control for Enhanced...
giuvecchio/matfuse
MatFuse: Controllable Material Generation with Diffusion Models (CVPR2024)
Shilin-LU/TF-ICON
[ICCV 2023] "TF-ICON: Diffusion-Based Training-Free Cross-Domain Image Composition" (Official...