Research map / Video / Video generation
Unwarping the Lens: A Physics-Grounded Approach to Video Glasses Removal
Radim Špetlík, David Futschik, Radek Danecek, Feitong Tan, Ziqian Bai, Rohit Pandey, Yinda Zhang
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 skip connection (CNN-CNN) (ablation) | CNN-CNN restoration net · FFHQ clear glasses | FID: 0.493 -> 0.439 | Improved |
| add global skip connection (ViT-CNN) (ablation) | JFSnet · FFHQ clear glasses | FID: 0.387 -> 0.379 | Improved |
| unfreeze last four DINOv2 layers (ViT-ViT, no skip) (ablation) | ViT-ViT · FFHQ clear glasses | FID: 1.131 -> 0.572 | Improved |
| unfreeze last four DINOv2 layers (ViT-ViT, with skip) (ablation) | ViT-ViT · FFHQ clear glasses | FID: 0.384 -> 0.392 | Hurt |
| use CNN decoder instead of ViT decoder (ablation) | JFSnet · FFHQ clear glasses | FID: 0.392 -> 0.379 | Improved |
| use ViT encoder with CNN decoder instead of CNN-only (ablation) | JFSnet · FFHQ clear glasses | FID: 0.439 -> 0.379 | Improved |
| add physics-based refraction and reflection simulation (ablation) | JFSnet · FFHQ clear glasses | FID | Improved |
| add translation-equivariance temporal loss (ablation) | JFSnet · CelebV-Text | RAFT-L1: 8.023 -> 7.682 | Improved |
| add translation-equivariance temporal loss (ablation) | JFSnet · CelebV-Text | tLPIPS: 19.520 -> 18.528 | Improved |
| add translation-equivariance temporal loss (ablation) | JFSnet · 300-identity paired set | SSIM: 0.947 -> 0.950 | No clear effect |
| retrain TOE on JFSnet training data | Take-Off-Eyeglasses · 300-identity paired set, lens region | PSNR: 18.98 -> 24.19 | Improved |
| retrain ProPainter on JFSnet training data | ProPainter · 300-identity paired set, lens region | PSNR: 15.28 -> 16.71 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- ProPainter — ProPainter: Improving Propagation and Transformer for Video Inpainting (ICCV 2023) · better on 3 of 3 shared results
- RAVE — RAVE: Randomized Noise Shuffling for Fast and Consistent Video Editing with Diffusion Models (CVPR 2024) · better on 1 of 1 shared result
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
- Diff-SemiER: Transparency-Aware Adaptive Fusion Diffusion Model with Generative Prior for Semi-Transparent Eyeglasses Removal (CVPR 2026 · similarity 0.90)
- TIGER: Taming Identity, Geometry, and Generative Priors for High-Quality Face Video Restoration (ECCV 2026 · similarity 0.89)
- OSDFace: One-Step Diffusion Model for Face Restoration (CVPR 2025 · similarity 0.87)
- Blind Face Restoration under Extreme Conditions: Leveraging 3D-2D Prior Fusion for Superior Structural and Texture Recovery (AAAI 2024 · similarity 0.86)
- Overcoming False Illusions in Real-World Face Restoration with Multi-Modal Guided Diffusion Model (ICLR 2025 · similarity 0.86)
- $InterLCM$: Low-Quality Images as Intermediate States of Latent Consistency Models for Effective Blind Face Restoration (ICLR 2025 · 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.
