Research map / Image generation / Image restoration
No Fear of the Dark: Image Retrieval Under Varying Illumination Conditions
Tomas Jenicek, Ondrej Chum
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 CLAHE photometric normalisation (ablation) | VGG GeM N/D · Tokyo 24/7 | mAP: 83.5 -> 87.0 | Improved |
| add jointly trained U-Net normalisation (ablation) | VGG GeM N/D · Tokyo 24/7 | mAP: 83.5 -> 86.5 | Improved |
| fine-tune on Retrieval-SfM-N/D day-night pairs (ablation) | VGG GeM · Tokyo 24/7 | mAP: 79.4 -> 83.5 | Improved |
| fine-tune on Retrieval-SfM-N/D day-night pairs (ablation) | CLAHE · Tokyo 24/7 | mAP: 84.1 -> 87.0 | Improved |
| fine-tune on Retrieval-SfM-N/D day-night pairs (ablation) | U-Net jointly · Tokyo 24/7 | mAP: 79.8 -> 86.5 | Improved |
| fine-tune on Retrieval-SfM-N/D day-night pairs (ablation) | VGG GeM · Revisited Oxford Medium | mAP: 60.9 -> 60.0 | Hurt |
| replace learned U-Net with CLAHE normalisation (ablation) | normalisation + VGG GeM N/D · Tokyo 24/7 | mAP: 86.5 -> 87.0 | No clear effect |
| fine-tune only the normalisation network (frozen embedding) (ablation) | U-Net N/D · Tokyo 24/7 | mAP: 86.4 -> 83.2 | Hurt |
| concatenate EdgeMAC descriptors with VGG GeM (ablation) | VGG GeM · Tokyo 24/7 | mAP: 79.4 -> 85.4 | Improved |
| concatenate EdgeMAC descriptors with VGG GeM (ablation) | VGG GeM · Revisited Oxford Medium | mAP: 60.9 -> 59.4 | Hurt |
| add CLAHE normalisation to the EdgeMAC+VGG N/D ensemble (ablation) | Edg+VGG N/D (1024D) · Tokyo 24/7 | mAP: 88.3 -> 90.5 | Improved |
| add CLAHE N/D normalisation (ablation) | VGG GeM · Tokyo 24/7 night query to day database | mAP: 67.3 -> 81.3 | Improved |
Closest papers by meaning
- Dark Side Augmentation: Generating Diverse Night Examples for Metric Learning (ICCV 2023 · similarity 0.86)
- Guided Curriculum Model Adaptation and Uncertainty-Aware Evaluation for Semantic Nighttime Image Segmentation (ICCV 2019 · similarity 0.86)
- Similarity Min-Max: Zero-Shot Day-Night Domain Adaptation (ICCV 2023 · similarity 0.85)
- See through the Dark: Learning Illumination-affined Representations for Nighttime Occupancy Prediction (NeurIPS 2025 · similarity 0.85)
- NightCC: Nighttime Color Constancy via Adaptive Channel Masking (CVPR 2024 · similarity 0.84)
- Boosting Object Detection with Zero-Shot Day-Night Domain Adaptation (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.
