Research map / Image generation / Super-resolution
SROBB: Targeted Perceptual Loss for Single Image Super-Resolution
Mohammad Saeed Rad, Behzad Bozorgtabar, Urs-Viktor Marti, Max Basler, Hazim Kemal Ekenel, Jean-Philippe Thiran
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 |
|---|---|---|---|
| replace standard perceptual loss with targeted perceptual loss (ablation) | SRGAN-style decoder with MSE and adversarial loss · Ablation user study, 35 images | user votes: 417 -> 1212 | Improved |
| add targeted perceptual loss on top of MSE-only training (ablation) | SR decoder (MSE + adversarial) · Ablation user study, 35 images | user votes: 49 -> 1212 | Improved |
| targeted perceptual loss wins images by majority vote versus others (ablation) | SR decoder · Ablation user study, 35 images | images won by majority: 30 of 35 vs 5 | Improved |
| use pixel-wise loss only on sky-dominated images (ablation) | SR decoder (MSE only) · Ablation user study, sky images | user votes | Mixed |
| use SROBB instead of ESRGAN | SR methods · User study, 35 COCO-Stuff images | user votes: 436 -> 617 | Improved |
| use SROBB instead of SFT-GAN (without segmentation map at test time) | SR methods · User study, 35 COCO-Stuff images | user votes: 223 -> 617 | Improved |
| use SROBB instead of SRGAN | SR methods · User study, 35 COCO-Stuff images | user votes: 201 -> 617 | Improved |
| use SROBB instead of RCAN | SR methods · User study, 35 COCO-Stuff images | user votes: 33 -> 617 | Improved |
| use SROBB instead of SRGAN | SR methods · Set5 baby | LPIPS: 0.112 -> 0.104 | Improved |
| use SROBB instead of SRGAN | SR methods · Set14 baboon | LPIPS: 0.220 -> 0.245 | Hurt |
| use SROBB instead of SRGAN | SR methods · Set5 baby | PSNR: 28.413 -> 28.869 | Improved |
| use SROBB instead of LapSRN | SR methods · Set5 baby / Set14 baboon | PSNR: 32.019 -> 28.869 | Hurt |
Closest papers by meaning
- Perception-Oriented Single Image Super-Resolution Using Optimal Objective Estimation (CVPR 2023 · similarity 0.90)
- Auto-Encoded Supervision for Perceptual Image Super-Resolution (CVPR 2025 · similarity 0.89)
- Beyond Image Super-Resolution for Image Recognition with Task-Driven Perceptual Loss (CVPR 2024 · similarity 0.89)
- Perceptual Losses for Real-Time Style Transfer and Super-Resolution (ECCV 2016 · similarity 0.88)
- Training Generative Image Super-Resolution Models by Wavelet-Domain Losses Enables Better Control of Artifacts (CVPR 2024 · similarity 0.86)
- Augmenting Perceptual Super-Resolution via Image Quality Predictors (CVPR 2025 · 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.
