Research map / 3D vision / Depth and stereo
Cascading Convolutional Color Constancy
Huanglin Yu, Ke Chen, Kaiqi Wang, Yanlin Qian, Zhaoxiang Zhang, Kui Jia
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 |
|---|---|---|---|
| Apply a 3-stage cascade with dependent illumination hypotheses | SqueezeNet-FC⁴ · Color Checker | mean angular error: 1.65 -> 1.35 | Improved |
| Apply a 3-stage cascade with dependent illumination hypotheses | SqueezeNet-FC⁴ · NUS 8-Camera | mean angular error: 2.23 -> 1.96 | Improved |
| Use the cascade instead of FFCC | Color Checker color constancy · Color Checker | mean angular error: 1.78 -> 1.35 | Improved |
| Use the cascade instead of FFCC | NUS 8-Camera color constancy · NUS 8-Camera | mean angular error: 1.99 -> 1.96 | Improved |
| Train and test across disjoint camera datasets | C⁴-SqueezeNet-FC⁴ · Camera-agnostic evaluation | worst 25% angular error | Improved |
| Supervise intermediate hypotheses as well as the final prediction (ablation) | Three-stage C⁴-SqueezeNet-FC⁴ · Color Checker | mean angular error: 1.48 -> 1.35 | Improved |
| Use equal loss weights rather than the other tested weight combinations (ablation) | Three-stage C⁴ · Color Checker | mean angular error: 1.37, 1.38, 1.37 -> 1.35 | Improved |
| Increase cascade stages from one to two (ablation) | C⁴ · Color Checker | angular error | Improved |
| Increase cascade stages beyond four (ablation) | C⁴ · Color Checker | angular error | No clear effect |
| Reduce backbone parameters by one third (ablation) | AlexNet-FC⁴ · Color Checker | mean angular error: 1.77 -> 2.17 | Hurt |
| Reduce backbone parameters by one third (ablation) | SqueezeNet-FC⁴ · Color Checker | mean angular error: 1.65 -> 1.94 | Hurt |
| Add a 3-stage cascade using a compressed AlexNet-FC⁴ backbone (ablation) | AlexNet-FC⁴, one-third parameters · Color Checker | mean angular error: 2.17 -> 1.65 | Improved |
Closest papers by meaning
- FC4: Fully Convolutional Color Constancy With Confidence-Weighted Pooling (CVPR 2017 · similarity 0.88)
- CLCC: Contrastive Learning for Color Constancy (CVPR 2021 · similarity 0.88)
- A Multi-Hypothesis Approach to Color Constancy (CVPR 2020 · similarity 0.87)
- GCC: Generative Color Constancy via Diffusing a Color Checker (CVPR 2025 · similarity 0.87)
- White-Balance First, Adjust Later: Cross-Camera Color Constancy via Vision-Language Evaluation (CVPR 2026 · similarity 0.84)
- NightCC: Nighttime Color Constancy via Adaptive Channel Masking (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.
