Research map / Image generation / Editing and style transfer
ReFocusEraser: Refocusing for Small Object Removal with Robust Context-Shadow Repair
Qingping Zheng, Bo Huang, Yang Liu, Haoyu Zhao, Ling Zheng, Zengmao Wang, Ying Li, Jiankang Deng
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 Camera-Adaptive Refocus (3x zoom, box stitching) to LoRA baseline (ablation) | FLUX.1-dev LoRA inpainter · RORD-val (small objects) | PSNR: 23.683 -> 31.082 | Improved |
| increase training focal length from 1x to 2x (ablation) | Stage-I inpainter · RORD-val (small objects) | FID: 40.287 -> 19.597 | Improved |
| increase training focal length from 1x to 3x (ablation) | Stage-I inpainter · RORD-val (small objects) | PSNR: 23.683 -> 29.574 | Improved |
| increase training focal length from 1x to 3x (ablation) | Stage-I inpainter · RORD-val (small objects) | FID: 40.287 -> 13.283 | Improved |
| use mask-based instead of box-based stitching at 2x (ablation) | Stage-I inpainter · RORD-val (small objects) | PSNR: 31.245 -> 31.905 | Improved |
| use mask-based instead of box-based stitching at 2x (ablation) | Stage-I inpainter · RORD-val (small objects) | FID: 8.458 -> 13.041 | Hurt |
| use mask-based instead of box-based stitching at 3x (ablation) | Stage-I inpainter · RORD-val (small objects) | PSNR: 31.082 -> 31.732 | Improved |
| use mask-based instead of box-based stitching at 3x (ablation) | Stage-I inpainter · RORD-val (small objects) | FID: 24.288 -> 25.036 | Hurt |
| add seam- and shadow-aware decoder repair on top of mask stitching (ablation) | ReFocusEraser · RORD-val (small objects) | FID: 25.036 -> 21.378 | Improved |
| replace L1 decoder loss with LPIPS (ablation) | Stage-II decoder · RORD-val (small objects) | FID: 24.580 -> 24.618 | No clear effect |
| L1 on background with LPIPS on foreground instead of L1 only (ablation) | Stage-II decoder · RORD-val (small objects) | FID: 24.580 -> 24.124 | No clear effect |
| LPIPS on background plus L1 on foreground instead of L1 only (ablation) | Stage-II decoder · RORD-val (small objects) | FID: 24.580 -> 21.379 | Improved |
Closest papers by meaning
- SmartEraser: Remove Anything from Images using Masked-Region Guidance (CVPR 2025 · similarity 0.87)
- ReFIR: Grounding Large Restoration Models with Retrieval Augmentation (NeurIPS 2024 · similarity 0.87)
- Towards Enhanced Image Inpainting: Mitigating Unwanted Object Insertion and Preserving Color Consistency (CVPR 2025 · similarity 0.86)
- InstructRestore: Region-Customized Image Restoration with Human Instructions (NeurIPS 2025 · similarity 0.86)
- Detail-Preserving Latent Diffusion for Stable Shadow Removal (CVPR 2025 · similarity 0.86)
- Paint by Inpaint: Learning to Add Image Objects by Removing Them First (CVPR 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.
