Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Multi-Domain Learning for Accurate and Few-Shot Color Constancy
Jin Xiao, Shuhang Gu, Lei Zhang
Check 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 |
|---|---|---|---|
| Share feature extraction and illuminant estimation across cameras, while using camera-specific channel re-weighting (ablation) | FC4-based color constancy · Gehler-Shi | mean angular error (°): 1.66 -> 1.59 | Improved |
| Share feature extraction and illuminant estimation across cameras, while using camera-specific channel re-weighting (ablation) | FC4-based color constancy · NUS-C600D | mean angular error (°): 1.97 -> 1.65 | Improved |
| Combine device datasets without camera-specific channel re-weighting (ablation) | FC4-based color constancy · Gehler-Shi | mean angular error (°): 1.66 -> 1.87 | Hurt |
| Combine device datasets without camera-specific channel re-weighting (ablation) | FC4-based color constancy · Cube+ | mean angular error (°): 1.35 -> 1.36 | No clear effect |
| Add camera-specific channel re-weighting to the shared model (ablation) | Multi-device Combination · Cube+ | mean angular error (°): 1.36 -> 1.24 | Improved |
| Increase the number of devices used for multi-domain training (ablation) | MDLCC · NUS-C600D | mean angular error (°): 1.82 -> 1.65 | Improved |
| Increase few-shot adaptation data from K=1 to K=20 (ablation) | MDLCC · Cube few-shot | mean angular error (°): 2.02 -> 1.47 | Improved |
| Adapt only camera-specific parameters using few-shot samples instead of fine-tuning all network weights | Few-shot color constancy · Cube few-shot, K=10 | mean angular error (°): 1.63 -> 1.56 | Improved |
| Adapt only camera-specific parameters using few-shot samples instead of fine-tuning all network weights | Few-shot color constancy · Gehler-Shi few-shot, K=10 | mean angular error (°): 2.66 -> 2.32 | Improved |
Closest papers by meaning
- FC4: Fully Convolutional Color Constancy With Confidence-Weighted Pooling (CVPR 2017 · similarity 0.90)
- Accelerated Self-Supervised Multi-Illumination Color Constancy With Hybrid Knowledge Distillation (TPAMI 2025 · similarity 0.90)
- A Multi-Hypothesis Approach to Color Constancy (CVPR 2020 · similarity 0.90)
- White-Balance First, Adjust Later: Cross-Camera Color Constancy via Vision-Language Evaluation (CVPR 2026 · similarity 0.86)
- GCC: Generative Color Constancy via Diffusing a Color Checker (CVPR 2025 · similarity 0.85)
- NightCC: Nighttime Color Constancy via Adaptive Channel Masking (CVPR 2024 · similarity 0.84)
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.
