Papers › DifFace: Blind Face Restoration with Diffused Error Contraction

DifFace: Blind Face Restoration with Diffused Error Contraction

13 Dec 2022arXiv:2212.06512archive 2025-07-28

Zongsheng Yue, Chen Change Loy

While deep learning-based methods for blind face restoration have achieved unprecedented success, they still suffer from two major limitations. First, most of them deteriorate when facing complex degradations out of their training data. Second, these methods require multiple constraints, e.g., fidelity, perceptual, and adversarial losses, which require laborious hyper-parameter tuning to stabilize and balance their influences. In this work, we propose a novel method named DifFace that is capable of coping with unseen and complex degradations more gracefully without complicated loss designs. The key of our method is to establish a posterior distribution from the observed low-quality (LQ) image to its high-quality (HQ) counterpart. In particular, we design a transition distribution from the LQ image to the intermediate state of a pre-trained diffusion model and then gradually transmit from this intermediate state to the HQ target by recursively applying a pre-trained diffusion model. The transition distribution only relies on a restoration backbone that is trained with L₂ loss on some synthetic data, which favorably avoids the cumbersome training process in existing methods. Moreover, the transition distribution can contract the error of the restoration backbone and thus makes our method more robust to unknown degradations. Comprehensive experiments show that DifFace is superior to current state-of-the-art methods, especially in cases with severe degradations. Code and model are available at https://github.com/zsyOAOA/DifFace.

PaperPDFCode

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

Code

zsyoaoa/difface officialmentioned in papermentioned on GitHubpytorchNOASSERTION report
zsyoaoa/resshift mentioned on GitHubpytorchNOASSERTION 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 Restoration

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Blind Face Restoration CelebA-Test DifFace FID 38.43 #5 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test DifFace IDS 62.39 #5 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test DifFace LPIPS 43.5 #5 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test DifFace PSNR 24.08 #5 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test DifFace SSIM 0.703 #5 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test GLEAN FID 60.3 #8 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test GLEAN IDS 67.13 #8 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test GLEAN LPIPS 46.9 #8 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test GLEAN PSNR 23.41 #8 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test GLEAN SSIM 0.666 #8 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test VQFR FID 45.84 #9 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test VQFR IDS 65.87 #9 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test VQFR LPIPS 47.1 #9 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test VQFR PSNR 21.94 #9 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test VQFR SSIM 0.585 #9 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test GFPGAN FID 46.99 #11 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test GFPGAN IDS 66.76 #11 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test GFPGAN LPIPS 49.5 #11 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test GFPGAN PSNR 22.18 #11 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test GFPGAN SSIM 0.631 #11 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test PSFRGAN FID 52.14 #12 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test PSFRGAN IDS 68.14 #12 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test PSFRGAN LPIPS 50 #12 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test PSFRGAN PSNR 22.74 #12 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test PSFRGAN SSIM 0.63 #12 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test PULSE FID 48.33 #13 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test PULSE IDS 74.97 #13 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test PULSE LPIPS 50.8 #13 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test PULSE PSNR 22.14 #13 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test PULSE SSIM 0.682 #13 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test DFDNet FID 64.65 #14 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test DFDNet IDS 86.21 #14 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test DFDNet LPIPS 55.4 #14 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test DFDNet PSNR 23.15 #14 of 15 Archive leaderboard report
Blind Face Restoration CelebA-Test DFDNet SSIM 0.629 #14 of 15 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

Diffusion

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