Research map / Image generation / Image restoration
GamutMLP: A Lightweight MLP for Color Loss Recovery
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use per-image optimized GamutMLP instead of clipping | GamutMLP · Gamut recovery, 200-image test set | RMSE: 0.0069 -> 0.0021 | Helped |
| Use per-image optimized GamutMLP instead of GamutNet | GamutMLP · Gamut recovery, 200-image test set | RMSE: 0.0052 -> 0.0021 | Helped |
| Use per-image optimized GamutMLP instead of ProPhoto-Sampled | GamutMLP · Gamut recovery, 200-image test set | RMSE: 0.0032 -> 0.0021 | Helped |
| Add meta-trained initialization to the encoded-input MLP | GamutMLP · Gamut recovery, 200-image test set | Optimization time: 16.32 sec -> 1.90 sec | Helped |
| Add meta-trained initialization to the encoded-input MLP | GamutMLP · Gamut recovery, 200-image test set | RMSE: 0.0021 -> 0.0021 | No clear effect |
| Use only spatial coordinates instead of both coordinates and color (ablation) | GamutMLP · Gamut recovery, 200-image test set | RMSE: 0.0021 -> 0.0037 | Hurt |
| Use only color values instead of both coordinates and color (ablation) | GamutMLP · Gamut recovery, 200-image test set | RMSE: 0.0021 -> 0.0028 | Hurt |
| Add both spatial coordinates and color values as MLP inputs (ablation) | GamutMLP · Gamut recovery, 200-image test set | RMSE: 0.0037 -> 0.0021; 0.0028 -> 0.0021 | Helped |
| Remove input feature encoding (ablation) | GamutMLP · Gamut recovery, 200-image test set | RMSE: 0.0021 -> 0.1467 | Hurt |
Closest papers by meaning
- GLUT: 3D Gaussian Lookup Table for Continuous Color Transformation (NeurIPS 2026 · similarity 0.86)
- Efficient Neural Network Encoding for 3D Color Lookup Tables (AAAI 2025 · similarity 0.86)
- Learning to Structure an Image With Few Colors (CVPR 2020 · similarity 0.86)
- Learning Color Equivariant Representations (ICLR 2025 · similarity 0.84)
- RGB-to-Polarization Estimation: A New Task and Benchmark Study (NeurIPS 2025 · similarity 0.83)
- PixelART: Image-to-Layer Decomposition without Latents or Text-to-Image Pretraining (NeurIPS 2026 · 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.