| license: cc-by-nc-sa-4.0 | |
| viewer: false | |
| task_categories: | |
| - image-to-image | |
| # NullFace: Training-Free Localized Face Anonymization | |
| [Hugging Face Paper](https://huggingface.co/papers/2503.08478) | |
| ## Test set | |
| We curated a subset of test subjects from the [CelebA-HQ](https://github.com/tkarras/progressive_growing_of_gans) and [FFHQ](https://github.com/NVlabs/ffhq-dataset) datasets for the quantitative comparisons against baseline methods in our paper. Specifically, we selected: | |
| - 4,852 images from [CelebA-HQ](https://github.com/tkarras/progressive_growing_of_gans) | |
| - 4,722 images from [FFHQ](https://github.com/NVlabs/ffhq-dataset) | |
| For each test subject, we generated a corresponding segmentation mask, which is designed to keep the eye and mouth areas visible when needed. | |
| The test subject lists (JSONL format) and segmentation masks can be downloaded in the following structure: | |
| ```bash | |
| nullface-test-set/ | |
| βββ celeba-hq/ | |
| β βββ mask_images/ | |
| β β βββ 00010.png | |
| β β βββ ... | |
| β βββ metadata.jsonl | |
| βββ ffhq/ | |
| β βββ mask_images/ | |
| β β βββ 00010.png | |
| β β βββ ... | |
| β βββ metadata.jsonl | |
| βββ README.md | |
| ``` |