shizongyan / lpff-dataset

(ICCV2023) A large-pose Flickr face dataset comprised of 19,590 high-quality real large-pose portrait images.

LPFF-dataset

LPFF is a large-pose Flickr face dataset comprised of 19,590 high-quality real large-pose portrait images.

[ICCV 2023] LPFF: A Portrait Dataset for Face Generators Across Large Poses

Yiqian Wu, Jing Zhang, Hongbo Fu, Xiaogang Jin*

The creation of 2D realistic facial images and 3D face shapes using generative networks has been a hot topic in recent years. Existing face generators exhibit exceptional performance on faces in small to medium poses (with respect to frontal faces), but struggle to produce realistic results for large poses. The distorted rendering results on large poses in 3D-aware generators further show that the generated 3D face shapes are far from the distribution of 3D faces in reality. We find that the above issues are caused by the training dataset's posture imbalance.

In this paper, we present LPFF, a large-pose Flickr face dataset comprised of 19,590 high-quality real large-pose portrait images. We utilize our dataset to train a 2D face generator that can process large-pose face images, as well as a 3D-aware generator that can generate realistic human face geometry. To better validate our pose-conditional 3D-aware generators, we develop a new FID measure to evaluate the 3D-level performance. Through this novel FID measure and other experiments, we show that LPFF can help 2D face generators extend their latent space and better manipulate the large-pose data, and help 3D-aware face generators achieve better view consistency and more realistic 3D reconstruction results.

Faces generated by EG3D models trained with FFHQ (Left) and Faces generated by EG3D models trained with our new dataset LPFF and FFHQ (Right).

News ✨ ✨

• [2023-07-14] Our paper is accepted by ICCV 2023!

• [2023-07-14] Our dataset, data processing code and training code are available now.

• [2023-03-28] The paper pdf is available: [2303.14407] LPFF: A Portrait Dataset for Face Generators Across Large Poses (arxiv.org)

• [2023-03-28] The supplementary pdf is available: Google Drive

Usage

Our dataset, code, and pretrained models are available now! ✨ ✨

For those who want to use our dataset, please first download the cropped images and predicted parameters, then add them to the FFHQ dataset, and finally perform dataset rebalance.

Available sources

Notice: We have uploaded all the data using OneDrive, and the shared link should be updated every two months. If you find that the link is not working, please contact us to update it.

	Description
dataset	Dataset download.
data_processing	Data processing codes and data download links. Including image alignment, camera parameters extraction, and dataset rebalance.
training	Model training and FID computation guidance.
networks	Pretrained StyleGAN2-ada and EG3D models trained on the LPFF+FFHQ dataset.

Contact

[email protected] / [email protected]

Citation

If you find this project helpful to your research, please consider citing:

@InProceedings{Wu_2023_ICCV,
    author    = {Wu, Yiqian and Zhang, Jing and Fu, Hongbo and Jin, Xiaogang},
    title     = {LPFF: A Portrait Dataset for Face Generators Across Large Poses},
    booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
    month     = {October},
    year      = {2023},
    pages     = {20327-20337}
}