Research map / Image generation / Image restoration
Multi-Scale Face Restoration With Sequential Gating Ensemble Network
Jianxin Lin, Tiankuang Zhou, Zhibo Chen
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 changed | Where | Measured | Effect |
|---|---|---|---|
| Add adversarial loss to MSE training (ablation) | SGEN · CelebA test, multiple image scales | MOS: 4.3667 -> 4.4833; 4.5000 -> 4.7833; 4.6000 -> 4.6333; 4.3667 -> 4.6333; 4.6333 -> 4.6000; 4.5833 -> 4.6833 | Improved |
| Add adversarial loss to MSE training (ablation) | SGEN · CelebA test, multiple image scales | PSNR: 23.00 -> 22.37; 23.60 -> 23.08; 24.12 -> 23.61; 24.61 -> 24.12; 24.98 -> 24.55; 25.39 -> 24.92 | Hurt |
| Replace sequential gating with max ensemble (ablation) | SGEN · CelebA test, multiple image scales | MOS: 4.4833 vs 2.0833; 4.7833 vs 2.1333; 4.6333 vs 2.0833; 4.6333 vs 1.9833; 4.6000 vs 2.1167; 4.6833 vs 2.2667 | Hurt |
| Replace sequential gating with average ensemble (ablation) | SGEN · CelebA test, multiple image scales | MOS: 4.4833 vs 3.2000; 4.7833 vs 3.4000; 4.6333 vs 3.2000; 4.6333 vs 3.2333; 4.6000 vs 3.3000; 4.6833 vs 3.2667 | Hurt |
| Replace sequential gating with concatenation ensemble (ablation) | SGEN · CelebA test, multiple image scales | MOS: 4.4833 vs 4.0333; 4.7833 vs 4.1000; 4.6333 vs 4.2000; 4.6333 vs 4.1000; 4.6000 vs 4.0500; 4.6833 vs 4.0833 | Hurt |
| Increase the number of levels from N=2 to N=3 (ablation) | SGEN · CelebA test | PSNR: 24.13 -> 24.28 | Improved |
| Increase the number of levels from N=3 to N=4 (ablation) | SGEN · CelebA test | PSNR: 24.28 -> 24.49 | Improved |
| Increase the number of levels from N=3 to N=4 (ablation) | SGEN · CelebA test | training time: 26h -> 35h | Hurt |
Closest papers by meaning
- EnsIR: An Ensemble Algorithm for Image Restoration via Gaussian Mixture Models (NeurIPS 2024 · similarity 0.85)
- Learning Warped Guidance for Blind Face Restoration (ECCV 2018 · similarity 0.85)
- A Recurrent Encoder-Decoder Network for Sequential Face Alignment (ECCV 2016 · similarity 0.84)
- Rethinking Deep Face Restoration (CVPR 2022 · similarity 0.84)
- RestoreFormer++: Towards Real-World Blind Face Restoration From Undegraded Key-Value Pairs (TPAMI 2023 · similarity 0.84)
- Omni-Kernel Network for Image Restoration (AAAI 2024 · similarity 0.83)
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.
