Research map / Detection and segmentation / Open-set and few-shot detection
Semantic-Guided Multi-Mask Image Harmonization
Xuqian Ren, Yifan Liu
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 |
|---|---|---|---|
| predict operator masks instead of regressing RGB directly (ablation) | Sg-MMH · HScene | MSE: 141.75 -> 99.84 | Improved |
| predict operator masks instead of regressing RGB directly (ablation) | Sg-MMH · HScene | PSNR: 28.73 -> 30.59 | Improved |
| predict operator masks instead of regressing RGB directly (ablation) | Sg-MMH · HLIP | MSE: 75.45 -> 48.02 | Improved |
| predict operator masks instead of regressing RGB directly (ablation) | Sg-MMH · HLIP | PSNR: 31.34 -> 33.11 | Improved |
| add patch discriminator with relativistic GAN loss to operator-mask generator (ablation) | Sg-MMH (OM) · HScene | MSE: 99.84 -> 94.00 | Improved |
| add patch discriminator with relativistic GAN loss to operator-mask generator (ablation) | Sg-MMH (OM) · HScene | PSNR: 30.59 -> 30.88 | Improved |
| add patch discriminator with relativistic GAN loss to operator-mask generator (ablation) | Sg-MMH (OM) · HLIP | MSE: 48.02 -> 42.62 | Improved |
| add patch discriminator with relativistic GAN loss to operator-mask generator (ablation) | Sg-MMH (OM) · HLIP | LPIPS: 0.021 -> 0.018 | Improved |
| predict operator masks instead of regressing RGB directly (ablation) | Sg-MMH · HScene | SSIM: 0.95 -> 0.96 | Improved |
| predict operator masks instead of regressing RGB directly (ablation) | Sg-MMH · HLIP | LPIPS: 0.025 -> 0.021 | Improved |
| direct RGB regression versus the composite input | Sg-MMH (RGB) · HLIP | LPIPS: 0.021 -> 0.025 | Hurt |
| use Sg-MMH(OMGAN) instead of DoveNet, BarginNet or RainNet on real composites | Sg-MMH (OMGAN) · Real composited images (user study) | user study score (%): 36.7 vs 20.6, 24.2, 18.5 | Improved |
Closest papers by meaning
- Training-and-Prompt-Free General Painterly Harmonization via Zero-Shot Disentenglement on Style and Content References (AAAI 2025 · similarity 0.88)
- FRIH: Fine-Grained Region-Aware Image Harmonization (AAAI 2024 · similarity 0.88)
- Progressive Painterly Image Harmonization from Low-Level Styles to High-Level Styles (AAAI 2024 · similarity 0.87)
- LEMaRT: Label-Efficient Masked Region Transform for Image Harmonization (CVPR 2023 · similarity 0.87)
- Semi-Supervised Parametric Real-World Image Harmonization (CVPR 2023 · similarity 0.87)
- Harmonizer: Learning to Perform White-Box Image and Video Harmonization (ECCV 2022 · similarity 0.87)
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.
