This dataset, randomly selected from MS-COCO, is used to fine-tune the target and shadow models in the NeurIPS 2024 Paper "Membership Inference on Text-to-Image Diffusion Models via Conditional Likelihood Discrepancy".
The dataset has been processed to conform to the format required by the GitHub repository CLiD code.
If you find it useful in your research, please consider citing our paper:
@article{zhai2024membership,
title={Membership… See the full description on the dataset page:
https://huggingface.co/datasets/zsf/COCO_MIA_ori_split1.
