Papers › HiFaceGAN: Face Renovation via Collaborative Suppression and Replenishment

HiFaceGAN: Face Renovation via Collaborative Suppression and Replenishment

11 May 2020arXiv:2005.05005archive 2025-07-28

Lingbo Yang, Chang Liu, Pan Wang, Shanshe Wang, Peiran Ren, Siwei Ma, Wen Gao

Existing face restoration researches typically relies on either the degradation prior or explicit guidance labels for training, which often results in limited generalization ability over real-world images with heterogeneous degradations and rich background contents. In this paper, we investigate the more challenging and practical "dual-blind" version of the problem by lifting the requirements on both types of prior, termed as "Face Renovation"(FR). Specifically, we formulated FR as a semantic-guided generation problem and tackle it with a collaborative suppression and replenishment (CSR) approach. This leads to HiFaceGAN, a multi-stage framework containing several nested CSR units that progressively replenish facial details based on the hierarchical semantic guidance extracted from the front-end content-adaptive suppression modules. Extensive experiments on both synthetic and real face images have verified the superior performance of HiFaceGAN over a wide range of challenging restoration subtasks, demonstrating its versatility, robustness and generalization ability towards real-world face processing applications.

PaperPDFCode

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

Lotayou/Face-Renovation officialmentioned in papermentioned on GitHubpytorchGPL-3.0 report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Blind Face RestorationFace HallucinationImage DenoisingImage RestorationImage Super-Resolution

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Blind Face Restoration CelebA-Test HiFaceGAN Deg. 42.18 #10 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test HiFaceGAN FID 66.09 #10 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test HiFaceGAN LPIPS 47.7 #10 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test HiFaceGAN NIQE 4.916 #10 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test HiFaceGAN PSNR 24.92 #10 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test HiFaceGAN SSIM 0.6195 #10 of 15 Archive leaderboard report
Face Hallucination FFHQ 512 x 512 - 16x upscaling HiFaceGAN FID 11.389 #1 of 4 Archive leaderboard report
Face Hallucination FFHQ 512 x 512 - 16x upscaling HiFaceGAN LPIPS 0.2449 #1 of 4 Archive leaderboard report
Face Hallucination FFHQ 512 x 512 - 16x upscaling HiFaceGAN NIQE 6.767 #1 of 4 Archive leaderboard report
Image Super-Resolution FFHQ 1024 x 1024 - 4x upscaling HiFaceGAN FID 1.978 #1 of 9 Archive leaderboard report
Image Super-Resolution FFHQ 1024 x 1024 - 4x upscaling HiFaceGAN MS-SSIM 0.975 #1 of 9 Archive leaderboard report
Image Super-Resolution FFHQ 1024 x 1024 - 4x upscaling HiFaceGAN PSNR 33.04 #1 of 9 Archive leaderboard report
Image Super-Resolution FFHQ 1024 x 1024 - 4x upscaling HiFaceGAN SSIM 0.875 #1 of 9 Archive leaderboard report
Image Super-Resolution FFHQ 256 x 256 - 4x upscaling HiFaceGAN FID 5.36 #1 of 11 Archive leaderboard report
Image Super-Resolution FFHQ 256 x 256 - 4x upscaling HiFaceGAN MS-SSIM 0.971 #1 of 11 Archive leaderboard report
Image Super-Resolution FFHQ 256 x 256 - 4x upscaling HiFaceGAN PSNR 28.65 #1 of 11 Archive leaderboard report
Image Super-Resolution FFHQ 256 x 256 - 4x upscaling HiFaceGAN SSIM 0.816 #1 of 11 Archive leaderboard report
Image Super-Resolution FFHQ 512 x 512 - 4x upscaling HiFaceGAN FED 0.0716 #1 of 8 Archive leaderboard report
Image Super-Resolution FFHQ 512 x 512 - 4x upscaling HiFaceGAN FID 1.898 #1 of 8 Archive leaderboard report
Image Super-Resolution FFHQ 512 x 512 - 4x upscaling HiFaceGAN LLE 2.071 #1 of 8 Archive leaderboard report
Image Super-Resolution FFHQ 512 x 512 - 4x upscaling HiFaceGAN LPIPS 0.0723 #1 of 8 Archive leaderboard report
Image Super-Resolution FFHQ 512 x 512 - 4x upscaling HiFaceGAN MS-SSIM 0.971 #1 of 8 Archive leaderboard report
Image Super-Resolution FFHQ 512 x 512 - 4x upscaling HiFaceGAN NIQE 6.961 #1 of 8 Archive leaderboard report
Image Super-Resolution FFHQ 512 x 512 - 4x upscaling HiFaceGAN PSNR 30.824 #1 of 8 Archive leaderboard report
Image Super-Resolution FFHQ 512 x 512 - 4x upscaling HiFaceGAN SSIM 0.838 #1 of 8 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

Adaptive Instance Normalization

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections