Awesome

MAGBIG

This is the repository for MAGBIG (Multilingual Assessment of Gender Bias in Image Generation) proposed in "Multilingual Text-to-Image Generation Magnifies Gender Stereotypes and Prompt Engineering May Not Help You".

You can find a Python package requirements file at translate_script\requirements.txt and our set of prompts for all prompt types and languages in the folder prompt.

Generating and Evaluating Images

You can generate images for these prompts using python generate_evaluate\generate_images.py. After that, you can classify them with python generate_evaluate\classify_images.py. For the evaluation, you can use python generate_evaluate\exp_max_unfairness.py for bias and python generate_evaluate\CLIPscore.py for text-to-image alignment with CLIP. These Python scripts also reproduce our results.

Generating and Translating prompts

To reproduce our prompts, you can run your bash script translate_script\run.sh or modify it to compute your own translations for new prompts and languages.

Please cite our work if you find it helpful.

Easy use of benchmark

You can also use our benchmark in the huggingface dataset library: https://huggingface.co/datasets/felfri/MAGBIG

citation

If you like or use our work, please consider citing us.

@misc{friedrich2024multilingual,
      title={Multilingual Text-to-Image Generation Magnifies Gender Stereotypes and Prompt Engineering May Not Help You}, 
      author={Felix Friedrich and Katharina Hämmerl and Patrick Schramowski and Jindrich Libovicky and Kristian Kersting and Alexander Fraser},
      year={2024},
      eprint={2401.16092},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}