Research map / Image generation / Image restoration
Conditional Sequential Modulation for Efficient Global Image Retouching
Jingwen He, Yihao Liu, Yu Qiao, Chao Dong
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 the condition network to the base network (ablation) | CSRNet base network · MIT-Adobe FiveK (expert C) | PSNR (dB): 20.47 -> 23.69 | Improved |
| Use 3x3 kernels instead of 1x1 in base network (with condition) (ablation) | CSRNet · MIT-Adobe FiveK (expert C) | PSNR (dB): 23.69 -> 23.73 | No clear effect |
| Increase base network from 3 to 7 layers (1x1) (ablation) | CSRNet · MIT-Adobe FiveK (expert C) | PSNR (dB): 23.69 -> 23.83 | Improved |
| Replace GFM with concatenation of condition vector (ablation) | CSRNet · MIT-Adobe FiveK (expert C) | PSNR (dB): 23.69 -> 23.31 | Hurt |
| Replace GFM with spatial SFTNet modulation (ablation) | CSRNet · MIT-Adobe FiveK (expert C) | PSNR (dB): 23.69 -> 20.73 | Hurt |
| Replace GFM with AdaFM 5x5 (ablation) | CSRNet · MIT-Adobe FiveK (expert C) | PSNR (dB): 23.69 -> 23.38 | Hurt |
| Replace GFM with AdaFM 3x3 (ablation) | CSRNet · MIT-Adobe FiveK (expert C) | PSNR (dB): 23.69 -> 23.70 | No clear effect |
| Concatenate brightness prior with learned condition vector (ablation) | CSRNet · MIT-Adobe FiveK (expert C) | PSNR (dB): 23.69 -> 23.01 | Hurt |
| Concatenate histograms prior with learned condition vector (ablation) | CSRNet · MIT-Adobe FiveK (expert C) | PSNR (dB): 23.69 -> 23.39 | Hurt |
| Add hand-crafted histogram prior to base network without condition network (ablation) | base network · MIT-Adobe FiveK (expert C) | PSNR (dB): 20.47 -> 22.90 | Improved |
| Finetune only the condition network for expert A instead of training from scratch (ablation) | CSRNet · MIT-Adobe FiveK (expert A) | PSNR (dB): 22.06 -> 22.29 | No clear effect |
| Finetune only the condition network for expert B instead of training from scratch (ablation) | CSRNet · MIT-Adobe FiveK (expert B) | PSNR (dB): 25.52 -> 25.61 | No clear effect |
Closest papers by meaning
- RetouchFormer: Semi-supervised High-Quality Face Retouching Transformer with Prior-Based Selective Self-Attention (AAAI 2024 · similarity 0.84)
- DiffRetouch: Using Diffusion to Retouch on the Shoulder of Experts (AAAI 2025 · similarity 0.84)
- VRetouchEr: Learning Cross-frame Feature Interdependence with Imperfection Flow for Face Retouching in Videos (CVPR 2024 · similarity 0.84)
- RetouchIQ: MLLM Agents for Instruction-Based Image Retouching with Generalist Reward (CVPR 2026 · similarity 0.84)
- RetouchGPT: LLM-based Interactive High-Fidelity Face Retouching via Imperfection Prompting (AAAI 2025 · similarity 0.84)
- Close Imitation of Expert Retouching for Black-and-White Photography (CVPR 2024 · similarity 0.83)
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.
