A-Debiasing-Framework-For-Attribute-Binding-In-Diffusion-Based-Text-To-Image

code release for paper:
"A Debiasing Framework for Attribute Binding in Diffusion-Based Text-to-Image"

📖 Overview

This repository provides the official implementation of our method proposed in the above paper.
We present a debiasing framework that enhances attribute-object binding in diffusion-based text-to-image generation models.

🛠️ Requirements

• Python 3.11.10
• PyTorch 2.4.1+cu121
• GPU: NVIDIA A800

The environment setup and part of the implementation are based on the following open-source projects:

• Magnet (I2-Multimedia-Lab)
• Rare-to-Frequent (Krafton AI)

🚀 Inference

To run inference using the provided script, first navigate to the script directory, then execute: bash inference.sh.

🙏 Acknowledgements

We sincerely thank the authors of Magnet and Rare-to-Frequent for their excellent open-source work, which served as the foundation and inspiration for this project.

📬 Contact

For questions, please contact: [email protected]