Awesome Face Restoration

A curated list of recources (papers, repositories etc.) about face restoration / face hallucination methods / face super-resolution methods. Feel free to contribute to this list.

What is face restoration and face hallucination?

Face hallucination refers to any super-resolution technique which applies specifically to faces. It comprises techniques which take noisy or low-resolution facial images, and convert them into high-resolution images using knowledge about typical facial features. From Wikipedia.

2022 Papers

Model name	Paper	Code
DMDNet	Learning Dual Memory Dictionaries for Blind Face Restoration	https://github.com/csxmli2016/DMDNet
RestoreFormer	RestoreFormer: High-Quality Blind Face Restoration from Undegraded Key-Value Pairs	https://github.com/wzhouxiff/RestoreFormer
CodeFormer	Towards Robust Blind Face Restoration with Codebook Lookup Transformer	https://github.com/sczhou/CodeFormer
GCFSR	GCFSR: a Generative and Controllable Face Super Resolution Method Without Facial and GAN Priors	https://github.com/hejingwenhejingwen/GCFSR
VQFR	VQFR: Blind Face Restoration with Vector-Quantized Dictionary and Parallel Decoder	https://github.com/TencentARC/VQFR
Panini-Net	Panini-Net: GAN Prior based Degradation-Aware Feature Interpolation for Face Restoration	https://github.com/jianzhangcs/panini

2021 Papers

Model name	Paper	Code
GFPGAN	Towards Real-World Blind Face Restoration with Generative Facial Prior	https://github.com/TencentARC/GFPGAN
GPEN	GAN Prior Embedded Network for Blind Face Restoration in the Wild	https://github.com/yangxy/GPEN
PSFR-GAN	Progressive Semantic-Aware Style Transformation for Blind Face Restoration	https://github.com/chaofengc/PSFRGAN
pixel2style2pixel (pSp)	Encoding in Style: a StyleGAN Encoder for Image-to-Image Translation	https://github.com/eladrich/pixel2style2pixel

2020 Papers

Model name	Paper	Code
DFDNet	Blind Face Restoration via Deep Multi-scale Component Dictionaries	https://github.com/csxmli2016/DFDNet
PULSE	PULSE: Self-Supervised Photo Upsampling via Latent Space Exploration of Generative Models	https://github.com/adamian98/pulse
HiFaceGAN	HiFaceGAN: Face Renovation via Collaborative Suppression and Replenishment	https://github.com/Lotayou/Face-Renovation
mGANprior	Image Processing Using Multi-Code GAN Prior	https://github.com/genforce/mganprior
UMSN-Face-Deblurring	Deblurring Face Images using Uncertainty Guided Multi-Stream Semantic Networks	https://github.com/rajeevyasarla/UMSN-Face-Deblurring

2019 Papers

Model name	Paper	Code
GWAInet	Exemplar Guided Face Image Super-Resolution without Facial Landmarks	https://github.com/berkdogan2/GWAInet

2018 Papers

Model name	Paper	Code
GFRNet	Learning Warped Guidance for Blind Face Restoration	https://github.com/csxmli2016/GFRNet

2017 Papers

Model name	Paper	Code
Super-FAN	Super-FAN: Integrated facial landmark localization and super-resolution of real-world low resolution faces in arbitrary poses with GANs	https://github.com/jzijin/Super-FAN
Wavelet-srnet	Wavelet-srnet: A wavelet-based cnn for multi-scale face super resolution	https://github.com/hhb072/WaveletSRNet

Paperswithcode links

• Image super-resolution
• Face hallucination
• Blind face restoration

Similiar resources

• Face-Hallucination-Benchmark

Image source: Towards Real-World Blind Face Restoration with Generative Facial Prior