Research map / Image generation / Image restoration
Diff-SemiER: Transparency-Aware Adaptive Fusion Diffusion Model with Generative Prior for Semi-Transparent Eyeglasses Removal
Jiahao Li, Shiqi Yin, Zhenxiang Lian, Jingtao Guo
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 restoration branch (TAFDM) with generative prior instead of GPDM alone (ablation) | Diff-SemiER · FFHQ-Test | IDD: 0.682 -> 0.534 | Improved |
| add generative prior from GPDM to TAFDM (ablation) | Diff-SemiER · FFHQ-Test | PSNR: 35.79 -> 36.16 | Improved |
| add generative prior from GPDM to TAFDM (ablation) | Diff-SemiER · FFHQ-Test | IDD: 0.581 -> 0.534 | Improved |
| add SMAF adaptive fusion on top of soft mask only (ablation) | Diff-SemiER · FFHQ-Test | PSNR: 35.64 -> 36.16 | Improved |
| add transparency soft mask modulation in SMAF (ablation) | Diff-SemiER · FFHQ-Test | PSNR: 36.10 -> 36.16 | Improved |
| add transparency soft mask modulation in SMAF (ablation) | Diff-SemiER · FFHQ-Test | IDD: 0.543 -> 0.534 | Improved |
| use Diff-SemiER instead of RDDM | semi-transparent eyeglasses removal · FFHQ-Test | PSNR: 35.33 -> 36.16 | Improved |
| use Diff-SemiER instead of RDDM on real-world CelebA-HQ | semi-transparent eyeglasses removal · Real-world CelebA-HQ | FID: 99.53 -> 99.80 | No clear effect |
| use Diff-SemiER instead of RDDM on real-world FFHQ | semi-transparent eyeglasses removal · Real-world FFHQ | FID: 66.21 -> 66.13 | No clear effect |
| use Diff-SemiER instead of RDDM in pairwise user study | semi-transparent eyeglasses removal · User study FFHQ | preference (%): 31.72%/68.28% | Improved |
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.
- Resfusion — Resfusion: Denoising Diffusion Probabilistic Models for Image Restoration Based on Prior Residual Noise (NeurIPS 2024) · better on 4 of 4 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
- Unwarping the Lens: A Physics-Grounded Approach to Video Glasses Removal (ECCV 2026 · similarity 0.90)
- Diff-Shadow: Global-guided Diffusion Model for Shadow Removal (AAAI 2025 · similarity 0.86)
- PHR-DIFF: Portrait Highlights Removal via Patch-aware Diffusion Model (AAAI 2025 · similarity 0.86)
- Selective Hourglass Mapping for Universal Image Restoration Based on Diffusion Model (CVPR 2024 · similarity 0.86)
- Diff-2-in-1: Bridging Generation and Dense Perception with Diffusion Models (ICLR 2025 · similarity 0.85)
- A Modular Conditional Diffusion Framework for Image Reconstruction (NeurIPS 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.
