Research map / Architectures and efficiency / CNN design
FC4: Fully Convolutional Color Constancy With Confidence-Weighted Pooling
Check 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 confidence-weighted pooling instead of average pooling on original images (ablation) | AlexNet-FC4 · Color Checker Dataset | mean angular error: 1.91 -> 1.77 | Helped |
| Use confidence-weighted pooling instead of average pooling on original images (ablation) | SqueezeNet-FC4 · Color Checker Dataset | mean angular error: 1.81 -> 1.65 | Helped |
| Use original-image input rather than resized input (ablation) | AlexNet-FC4 with weighted pooling · Color Checker Dataset | mean angular error: 1.89 -> 1.77 | Helped |
| Increase patch input from 1 to 8 patches (ablation) | AlexNet-FC4 with weighted pooling · Color Checker Dataset | mean angular error: 1.94 -> 1.83 | Helped |
| Use original-image input rather than 8-patch input (ablation) | AlexNet-FC4 with weighted pooling · Color Checker Dataset | evaluation time: 0.052 -> 0.025 | Helped |
| Use median rather than average pooling for patch estimates (ablation) | AlexNet-FC4 with weighted pooling and 8 patches · Color Checker Dataset | mean angular error: 1.83 -> 1.88 | Hurt |
| Remove conv4 and/or conv5 (ablation) | FC4 · Color Checker Dataset | angular error | Hurt |
| Replace the 6 × 6 conv6 kernel with a 1 × 1, 3 × 3, or 10 × 10 kernel (ablation) | FC4 · Color Checker Dataset | angular error | Hurt |
| Use resized rather than original-image input (ablation) | AlexNet-FC4 with average pooling · Color Checker Dataset | mean angular error: 1.91 -> 1.96 | Mixed |
| Use SqueezeNet rather than AlexNet as the FC4 backbone (ablation) | FC4 · Color Checker Dataset | mean angular error: 1.77 -> 1.65 | Helped |
| Use SqueezeNet rather than AlexNet as the FC4 backbone (ablation) | FC4 · NUS 8-Camera Dataset | mean angular error: 2.12 -> 2.23 | Hurt |
Closest papers by meaning
- GCC: Generative Color Constancy via Diffusing a Color Checker (CVPR 2025 · similarity 0.88)
- White-Balance First, Adjust Later: Cross-Camera Color Constancy via Vision-Language Evaluation (CVPR 2026 · similarity 0.87)
- NightCC: Nighttime Color Constancy via Adaptive Channel Masking (CVPR 2024 · similarity 0.85)
- Learning Color Equivariant Representations (ICLR 2025 · similarity 0.85)
- ChromaFusionNet (CFNet): Natural Fusion of Fine-Grained Color Editing (AAAI 2024 · similarity 0.84)
- Bright-NeRF: Brightening Neural Radiance Field with Color Restoration from Low-Light RAW Images (AAAI 2025 · 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.