Blind Face Restoration via Integrating Face Shape and Generative Priors

Zhu, Feida; Zhu, Jun-Wei; Chu, Wenqing; Zhang, Xinyi; Ji, Xiaozhong; Wang, Chengjie; Tai, Ying

doi:10.1109/cvpr52688.2022.00751

Cited by 18 publications

(9 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In GPEN, they first learn a GAN used for generating high-quality face images and embed this pre-trained GAN into a deep neural network as a decoder prior for face restoration. More recently, Zhu et al [100] propose to combine shape and generative prior to guide the process of face restoration for the network. In the proposed network, they first use a shape restoration module to generate shape prior.…”

Section: Generative Prior Based Deep Restoration Methodsmentioning

confidence: 99%

“…[23], [28], [99], [100], they the latent prior in pre-trained face GAN models such as StyleGAN [80] and incorporate the prior into the process of face restoration. One representative work is GFP-GAN which effectively leverages face priors encapsulated in the pre-trained face GAN to perform the face restoration.…”

Section: Datasetsmentioning

confidence: 99%

“…It can be seen that L2 loss is only sensitive to large errors, while L1 loss treats larger and small errors equally. Early methods [21], [77], [85] usually use L2 loss in their models and recent works [19], [88], [100] mainly resort to L1 loss. While the pixel-wise loss can force the model to achieve high PSNR values, it often results in over-smooth and unrealistic images [33], [85].…”

Section: Loss Functionsmentioning

confidence: 99%

“…MFPSNet [88], GWAInet [73], GFP-GAN [9], GPEN [99], SGPN [100], MSCNN [3], UMSN [4], RestoreFormer [19], FaceFormer [101],…”

Section: L1 Lossmentioning

confidence: 99%

“…GPEN [99], SGPN [100], MSCNN [3], UMSN [4], NPFNet [87], UR-DGN [18], MCGAN [94], HiFaceGAN [8], CodeFormer [102],…”

Section: L1 Lossmentioning

confidence: 99%

See 4 more Smart Citations

A Survey of Deep Face Restoration: Denoise, Super-Resolution, Deblur, Artifact Removal

Wang¹,

Chen²,

Luo³

et al. 2022

Preprint

View full text Add to dashboard Cite

Face Restoration (FR) aims to restore High-Quality (HQ) faces from Low-Quality (LQ) input images, which is a domain-specific image restoration problem in the low-level computer vision area. The early face restoration methods mainly use statistic priors and degradation models, which are difficult to meet the requirements of real-world applications in practice. In recent years, face restoration has witnessed great progress after stepping into the deep learning era. However, there are few works to study deep learning-based face restoration methods systematically. Thus, this paper comprehensively surveys recent advances in deep learning techniques for face restoration. Specifically, we first summarize different problem formulations and analyze the characteristic of the face image. Second, we discuss the challenges of face restoration. Concerning these challenges, we present a comprehensive review of existing FR methods, including prior based methods and deep learning-based methods. Then, we explore developed techniques in the task of FR covering network architectures, loss functions, and benchmark datasets. We also conduct a systematic benchmark evaluation on representative methods. Finally, we discuss future directions, including network designs, metrics, benchmark datasets, applications, etc. We also provide an open-source repository for all the discussed methods, which is available at https://github.com/TaoWangzj/Awesome-Face-Restoration.

show abstract

Section: Generative Prior Based Deep Restoration Methodsmentioning

confidence: 99%