Top: The architecture of OnmiFace.
Face restoration is a challenging task due to the need to remove artifacts and restore details. Traditional methods usually use generative model prior to achieve face restoration, but the restored results are still insufficient in terms of realism and details. In this paper, we introduce OmniFace, a novel face restoration framework that leverages Transformer-based diffusion flow. By exploiting the scaling property of Transformer, OmniFace achieves high-resolution restoration with exceptional realism and detail. The framework integrates three key components: (1) a Transformer-driven vector estimation network, (2) a representation aligned ControlNet, and (3) an adaptive training strategy for face restoration. The inherent scaling law of Transformer architectures enables the restoration of high-quality faces at high resolution. The controlnet combined with pre-trained diffusion representation can be easily trained. The adaptive training strategy provides a vector field that is more suitable for face restoration. Comprehensive experiments demonstrate that OmniFace outperforms existing techniques in terms of restoration quality across multiple benchmark datasets, especially in restoring photographic-level texture details in high-resolution scenes.
Quantitative Results in Val datasets.
Qualitative results. Even though input faces are severely degraded, our OnmiFace produces high-quality faces with faithful details.
We would like to thank Wei Chen for his important contribution to the experiment, Xu Li for her constructive suggestions on the thesis design, and He Gang for his support in the process of establishing the thesis topic.
OnmiFace: Unleashing the Potential of Transformer Flow for Photorealistic Face Restoration
@inproceedings{onmiface,
author = {Kepeng Xu and Li Xu and Gang He and Wei Chen and Xianyun Wu and Wenxin Yu},
title = {Unleashing the Potential of Transformer Flow for Photorealistic Face Restoration},
booktitle = {IJCAI 2025},
year = {2025},
}