Research map / Image generation / Image restoration
What Else Can Fool Deep Learning? Addressing Color Constancy Errors on Deep Neural Network Performance
Mahmoud Afifi, Michael S. Brown
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 WB augmentation to training (ablation) | SmallNet · CIFAR-10 (different WB test) | accuracy: 0.655 -> 0.786 | Improved |
| add WB augmentation to training (ablation) | AlexNet · CIFAR-10 (different WB test) | accuracy: 0.797 -> 0.889 | Improved |
| add WB augmentation to training (ablation) | AlexNet · CIFAR-100 (different WB test) | accuracy: 0.526 -> 0.670 | Improved |
| add WB augmentation to training (ablation) | AlexNet · CIFAR-100 (original test) | accuracy: 0.768 -> 0.735 | Hurt |
| add WB augmentation to training (ablation) | SmallNet · CIFAR-10 (original test) | accuracy: 0.799 -> 0.809 | Improved |
| add WB augmentation to training (ablation) | SegNet · ADE20K (different WB test) | IoU: 0.180 -> 0.199 | Improved |
| add WB augmentation to training (ablation) | SegNet · ADE20K (different WB test) | pixel accuracy: 0.557 -> 0.581 | Improved |
| add WB augmentation to training (ablation) | SegNet · ADE20K (original test) | pixel accuracy: 0.603 -> 0.597 | Hurt |
| add WB augmentation to training (ablation) | SmallNet · Cat-2 in-camera different WB | accuracy: 0.404 -> 0.496 | Improved |
| add WB augmentation to training (ablation) | AlexNet · Cat-2 in-camera different WB | accuracy: 0.734 -> 0.788 | Improved |
| use RGB augmentation instead of WB augmentation (ablation) | SegNet · ADE20K (original test) | pixel accuracy: 0.603 -> 0.544 | Hurt |
| train on white-balanced images instead of original (ablation) | AlexNet · CIFAR-10 (different WB test) | accuracy: 0.797 -> 0.811 | Improved |
Closest papers by meaning
- When Color Constancy Goes Wrong: Correcting Improperly White-Balanced Images (CVPR 2019 · similarity 0.90)
- Deep White-Balance Editing (CVPR 2020 · similarity 0.86)
- White-Balance First, Adjust Later: Cross-Camera Color Constancy via Vision-Language Evaluation (CVPR 2026 · similarity 0.86)
- FC4: Fully Convolutional Color Constancy With Confidence-Weighted Pooling (CVPR 2017 · similarity 0.86)
- Ranking-Based Color Constancy With Limited Training Samples (TPAMI 2023 · similarity 0.85)
- HVDualformer: Histogram-Vision Dual Transformer for White Balance (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.
