Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Similarity Min-Max: Zero-Shot Day-Night Domain Adaptation
Rundong Luo, Wenjing Wang, Wenhan Yang, Jiaying Liu
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 |
|---|---|---|---|
| train D then F in two steps instead of alternating like GAN (ablation) | similarity min-max pipeline · CODaN nighttime test | top-1 accuracy (%): 63.84 -> 65.87 | Improved |
| use learned exposure-guided quadratic-curve darkening instead of brightness adjustment (ablation) | darkening module D · CODaN nighttime test | top-1 accuracy (%): 57.96 -> 65.87 | Improved |
| use learned darkening instead of fixed gamma correction (ablation) | darkening module D · CODaN nighttime test | top-1 accuracy (%): 63.96 -> 65.87 | Improved |
| use iterative quadratic curve instead of reciprocal curve (ablation) | darkening module D · CODaN nighttime test | top-1 accuracy (%): 62.60 -> 65.87 | Improved |
| use iterative quadratic curve instead of gamma curve (ablation) | darkening module D · CODaN nighttime test | top-1 accuracy (%): 64.16 -> 65.87 | Improved |
| add similarity minimization loss L_D^sim (ablation) | darkening module D · CODaN nighttime test | top-1 accuracy (%): 64.56 -> 65.87 | Improved |
| add similarity maximization loss L_F^sim (ablation) | feature extractor F · CODaN nighttime test | top-1 accuracy (%): 64.88 -> 65.87 | Improved |
| add both similarity losses (min and max) (ablation) | similarity min-max framework · CODaN nighttime test | top-1 accuracy (%): 64.08 -> 65.87 | Improved |
| add low-light enhancement as preprocessing for daytime baseline | RefineNet · Nighttime Driving | mIoU (%): 34.3 -> 25.1 (RUAS) | Hurt |
| apply the min-max framework to segmentation, place recognition and video | daytime baselines · Nighttime Driving, Dark-Zurich, Tokyo 24/7, ARID | task accuracy: ARID 47.02 -> 51.52; Dark-Zurich 30.6 -> 40.2 | Improved |
Closest papers by meaning
- Exploring the Common Appearance-Boundary Adaptation for Nighttime Optical Flow (ICLR 2024 · similarity 0.89)
- The Parables of the Mustard Seed and the Yeast: Extremely Low-Budget, High-Performance Nighttime Semantic Segmentation (AAAI 2025 · similarity 0.88)
- Boosting Object Detection with Zero-Shot Day-Night Domain Adaptation (CVPR 2024 · similarity 0.88)
- NightAdapter: Learning a Frequency Adapter for Generalizable Night-time Scene Segmentation (CVPR 2025 · similarity 0.87)
- NightRain: Nighttime Video Deraining via Adaptive-Rain-Removal and Adaptive-Correction (AAAI 2024 · similarity 0.86)
- NightCC: Nighttime Color Constancy via Adaptive Channel Masking (CVPR 2024 · similarity 0.86)
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.
