Text-to-Image Association Test (T2IAT)

This is the code implementation for Text-to-Image Association Test (T2IAT).

To run the script for image generations,

python3 txt2img.py

To run the association test with the image generations, go through the bias-test.ipynb jupyter notebook.

To deploy the gradio demo, run

gradio demo.py

and open localhost:7860 in your local web browser.

Requirements

pip3 install git+https://github.com/openai/CLIP.git
pip3 install --upgrade diffusers[torch]
pip3 install gradio  # only used for demo

Citation

@inproceedings{wang-etal-2022-assessing,
    title = "T2IAT: Measuring Valence and Stereotypical Biases in Text-to-Image Generation",
    author = "Wang, Jialu  and
      Liu, Xinyue Gabby  and
      Di, Zonglin  and
      Liu, Yang  and
      Wang, Xin",
    booktitle = "Findings of the Association for Computational Linguistics: ACL 2023",
    month = July,
    year = "2023",
    publisher = "Association for Computational Linguistics",
}