AI Paper Atlas

Research map / Image generation / Image restoration

RestoreFormer++: Towards Real-World Blind Face Restoration From Undegraded Key-Value Pairs

Zhouxia Wang, Jiawei Zhang, Tianshui Chen, Wenping Wang, Ping Luo

TPAMI 2023 · Image generation › Image restoration · approach: Priors for restoration networks · 43 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add ROHQD prior features instead of degraded only (ablation)RestoreFormer++ (MHSA) · CelebA-TestFID: 47.96 -> 42.53Improved
add ROHQD prior features instead of degraded only (ablation)RestoreFormer++ (MHSA) · CelebA-TestIDD: 0.6461 -> 0.7467Hurt
fuse degraded and prior with multi-head cross attention (MHCA-D) instead of SFT (ablation)RestoreFormer++ · CelebA-TestFID: 44.67 -> 42.25Improved
use MHCA-P instead of MHCA-D (ablation)RestoreFormer++ · CelebA-TestFID: 42.25 -> 39.31Improved
use two scales instead of one for fusion (ablation)RestoreFormer++ · CelebA-TestFID: 39.31 -> 38.41Improved
use three scales instead of two (ablation)RestoreFormer++ · CelebA-TestFID: 38.41 -> 39.11No clear effect
add extending degrading model (EDM) (ablation)RestoreFormer++ · WebPhoto-TestFID: 75.42 -> 74.21Improved
add EDM (ablation)RestoreFormer++ · LFW-TestFID: 48.10 -> 48.48No clear effect
add EDMGFP-GAN · WebPhoto-TestFID: 87.82 -> 86.17Improved
add identity loss (ablation)RestoreFormer++ · CelebA-TestIDD: 0.6052 -> 0.5375Improved
add perceptual loss (ablation)RestoreFormer++ · CelebA-TestFID: 48.66 -> 38.41Improved
add adversarial and component losses (ablation)RestoreFormer++ · CelebA-TestFID: 54.14 -> 38.41Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy