non-hair-FFHQ
The non-hair-FFHQ dataset is a high-quality image dataset that contains 6,000 non-hair FFHQ portraits, based on stylegan2-ada and ffhq-dataset.
The dataset is built by our HairMapper method.
HairMapper: Removing Hair from Portraits Using GANs
Yiqian Wu, Yongliang Yang, Xiaogang Jin*.
2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition.
[Paper (4.21MB)] [Video (46.7MB)] [Suppl (4.42M)] [Project Page] [code]
[Paper-high resolution (25.8MB)] [Suppl-high resolution (16.4M)]
We apply our method on FFHQ images (all images have licenses that allow free use, redistribution, and adaptation for non-commercial purposes) and present a non-hair-FFHQ dataset that contains 6,000 non-hair portraits to inspire and facilitate more works in the future.
Overview
Google drive link of the dataset : https://drive.google.com/drive/folders/1CbyFYDTUqWRneyuDlVznY4XG-8pLhoAS?usp=sharing.
| dir | information |
|---|---|
| ├ hair | original images, {img_id}.png |
| â”” non-hair | results images , {img_id}.png |
Code
https://github.com/oneThousand1000/HairMapper
Agreement
The non-hair-FFHQ dataset is available for non-commercial research purposes only.
Related Works
A Style-Based Generator Architecture for Generative Adversarial Networks Tero Karras (NVIDIA), Samuli Laine (NVIDIA), Timo Aila (NVIDIA) https://arxiv.org/abs/1812.04948
Training Generative Adversarial Networks with Limited Data Tero Karras, Miika Aittala, Janne Hellsten, Samuli Laine, Jaakko Lehtinen, Timo Aila https://arxiv.org/abs/2006.06676
Citation
@InProceedings{Wu_2022_CVPR,
author = {Wu, Yiqian and Yang, Yong-Liang and Jin, Xiaogang},
title = {HairMapper: Removing Hair From Portraits Using GANs},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2022},
pages = {4227-4236}
}