Research map / Video / Video segmentation and tracking
Recurrent Color Constancy
Yanlin Qian, Ke Chen, Jarno Nikkanen, Joni-Kristian Kamarainen, Jiri Matas
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 simulated zoom-in sequence (SS) (ablation) | RCC-Net · SFU Gray Ball (non-linear) | Median angular error (deg): 3.2 -> 2.9 | Improved |
| add simulated zoom-in sequence (SS) (ablation) | RCC-Net · SFU Gray Ball (linear) | Median angular error (deg): 6.3 -> 5.1 | Improved |
| use SS instead of random patches (ablation) | RCC-Net · SFU Gray Ball (linear) | Median angular error (deg): 4.9 -> 5.1 | Improved |
| use spatial pyramid regions instead of SS (ablation) | RCC-Net · SFU Gray Ball (non-linear) / SFU Gray Ball (linear) | Median angular error (deg): 2.9 / 4.7 vs SS 2.9 / 5.1 | No clear effect |
| add any region-generation augmentation (RP, SP, SS) vs none (ablation) | RCC-Net · non-linear/linear | Median angular error (deg): 3.2 / 6.3 -> 2.9 / 5.1 | Improved |
| increase subsequence length N from 1 to 5 (ablation) | RCC-Net · SFU Gray Ball (non-linear) | Median angular error (deg): 3.3 -> 2.9 | Improved |
| increase subsequence length N from 1 to 5 (ablation) | RCC-Net · SFU Gray Ball (linear) | Median angular error (deg): 5.5 -> 5.1 | Improved |
| increase subsequence length N from 5 to 10 (ablation) | RCC-Net · SFU Gray Ball (non-linear) / SFU Gray Ball (linear) | Median angular error (deg): 2.9 / 5.1 -> 2.9 / 5.2 | No clear effect |
| use angular loss instead of MSE loss (with SS) (ablation) | RCC-Net · SFU Gray Ball (non-linear) | Median angular error (deg): 3.8 -> 2.9 | Improved |
| use angular loss instead of MSE loss (with SS) (ablation) | RCC-Net · SFU Gray Ball (linear) | Median angular error (deg): 5.4 -> 5.1 | Improved |
| use angular loss instead of MSE loss (no SS) (ablation) | RCC-Net · SFU Gray Ball (linear) | Median angular error (deg): 6.0 -> 6.3 | Mixed |
| switch from CNN to convolutional LSTM (+LSTM) (ablation) | RCC-Net · SFU Gray Ball (linear) | Median angular error (deg): 17% lower median angular error | 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 0.88)
- Ranking-Based Color Constancy With Limited Training Samples (TPAMI 2023 · similarity 0.88)
- GCC: Generative Color Constancy via Diffusing a Color Checker (CVPR 2025 · similarity 0.86)
- NightCC: Nighttime Color Constancy via Adaptive Channel Masking (CVPR 2024 · similarity 0.85)
- White-Balance First, Adjust Later: Cross-Camera Color Constancy via Vision-Language Evaluation (CVPR 2026 · similarity 0.85)
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.
