Research map / Image generation / Image restoration
Deep Semantic Face Deblurring
Ziyi Shen, Wei-Sheng Lai, Tingfa Xu, Jan Kautz, Ming-Hsuan Yang
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 global semantic priors to content-loss training (ablation) | Face deblurring CNN · Helen and CelebA test sets | PSNR: 24.85 -> 25.32; 24.23 -> 24.32 | Improved |
| add global semantic priors to content-loss training (ablation) | Face deblurring CNN · CelebA test set | SSIM: 0.864 -> 0.864 | No clear effect |
| add local structural loss to global semantic priors (ablation) | Face deblurring CNN · Helen and CelebA test sets | PSNR: 25.32 -> 25.48; 24.32 -> 24.58 | Improved |
| add incremental training (ablation) | Face deblurring CNN · Helen and CelebA test sets | PSNR: 25.48 -> 25.55; 24.58 -> 24.61 | Improved |
| replace direct training with incremental training (ablation) | Face deblurring CNN · Helen test set, different blur-kernel sizes | PSNR | Improved |
| add perceptual loss (ablation) | Face deblurring CNN with incremental training · Helen and CelebA test sets | PSNR: 25.55 -> 25.99; 24.61 -> 25.05 | Improved |
| add adversarial loss after perceptual loss (ablation) | Face deblurring CNN · Helen and CelebA test sets | PSNR: 25.99 -> 25.58; 25.05 -> 24.34 | Hurt |
| add adversarial loss after perceptual loss (ablation) | Face deblurring CNN · Helen and CelebA test sets | photorealism | Improved |
| use semantic priors and local structural losses | Face deblurring CNN · CelebA test set | top-1 recognition accuracy: 42\% -> 54\% | Improved |
| use semantic priors and local structural losses | Face deblurring CNN · CelebA test set | face detection success rate: 95\% -> 99\% | Improved |
| use the proposed face deblurring method instead of Nah et al. | Face deblurring · Helen and CelebA test sets | PSNR: 24.12 -> 25.58; 22.43 -> 24.34 | Improved |
| use the proposed face deblurring method instead of Nah et al. | Face deblurring · 128 × 128 images | execution time (seconds): 0.09 -> 0.05 | Improved |
Closest papers by meaning
- Exploiting Semantics for Face Image Deblurring (IJCV 2020 · similarity 0.98)
- Deep Image Deblurring: A Survey (IJCV 2022 · similarity 0.92)
- Learning to Deblur Face Images via Sketch Synthesis (AAAI 2020 · similarity 0.92)
- Learning to Deblur Images with Exemplars (TPAMI 2018 · similarity 0.91)
- Joint Face Hallucination and Deblurring via Structure Generation and Detail Enhancement (IJCV 2019 · similarity 0.91)
- Learning Degradation-unaware Representation with Prior-based Latent Transformations for Blind Face Restoration (CVPR 2024 · similarity 0.86)
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.
