Research map / 3D vision / Face and shape
Motion Deblurring of Faces
Grigorios G. Chrysos, Paolo Favaro, Stefanos P. Zafeiriou
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 HG conditioning to the blur-removal generator (ablation) | face deblurring GAN · 300VW (PrAvg 7) | AUC: 0.759 -> 0.783 | Improved |
| add HG conditioning to the generator (ablation) | face deblurring GAN · 300VW (PrAvg 7) | failure rate (%): 7.895 -> 5.962 | Improved |
| add additional label data (2MF2 backward data) to training (ablation) | identity-noHG · 300VW (PrAvg 7) | AUC: 0.751 -> 0.759 | Improved |
| add the identity branch (ablation) | face deblurring GAN · 300VW (PrAvg 7) | AUC: 0.783 -> 0.798 | Improved |
| add the identity branch (ablation) | face deblurring GAN · 300VW (PrAvg 7) | failure rate (%): 5.962 -> 3.947 | Improved |
| add the identity branch (ablation) | face deblurring GAN · 300VW (PrAvg 7) | SSIM: 0.92883 -> 0.93501 | Improved |
| deblur with the proposed method before landmark localization instead of leaving blurred | landmark localization · 300VW (PrAvg 7) | AUC: 0.743 -> 0.798 | Improved |
| use ours instead of Nah et al (2017) | landmark localization · 300VW (PrAvg 7) | AUC: 0.792 -> 0.798 | Improved |
| use ours instead of Nah et al (2017) under non-averaging VLA blur | landmark localization · 300VW (VLA) | AUC: 0.767 -> 0.786 | Improved |
| increase averaged frames from 7 to 21 | proposed method · 300VW | AUC: 0.798 -> 0.748 | Hurt |
| use optimization-based deblurring (Xu et al, Chakrabarti) instead of learning-based | landmark localization · 300VW (PrAvg 15) | AUC: 0.446 and 0.437 vs 0.769 | Hurt |
| deblur with Nah et al (2017) before localization instead of blurred input | landmark localization · 300VW (PrAvg 21) | failure rate (%): 9.270 -> 7.255 | Improved |
Closest papers by meaning
- Face Video Deblurring Using 3D Facial Priors (ICCV 2019 · similarity 0.88)
- FAB: A Robust Facial Landmark Detection Framework for Motion-Blurred Videos (ICCV 2019 · similarity 0.87)
- Learning to Deblur Face Images via Sketch Synthesis (AAAI 2020 · similarity 0.86)
- ID-Blau: Image Deblurring by Implicit Diffusion-based reBLurring AUgmentation (CVPR 2024 · similarity 0.86)
- Parameterized Blur Kernel Prior Learning for Local Motion Deblurring (CVPR 2025 · similarity 0.85)
- Motion-adaptive Separable Collaborative Filters for Blind Motion Deblurring (CVPR 2024 · similarity 0.85)
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.
