AI Paper Atlas

Research map / Label-efficient and robust learning / Domain adaptation and generalization

Towards Generalized Representations for Low-Light Understanding: When Signal Constancy Meets Semantic Enrichment

Yifan Li, Haofeng Huang, Wenhan Yang, Jiaying Liu

CVPR 2026 · Label-efficient and robust learning › Domain adaptation and generalization · approach: Augmentation and mixture-of-experts generalization

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 changedWhereMeasuredEffect
add illumination-invariant prior and VFM semantic enrichment (UniPrior) without real low-light dataResNet-18 · CODaN classificationTop-1 accuracy (%): 53.32 -> 74.52Improved
add illumination-invariant prior as auxiliary input (ablation)UniPrior · CODaN classificationTop-1 accuracy (%): 67.24 -> 74.52Improved
zero-initialize the prior channels in the merge convolution (ablation)UniPrior · CODaN classificationTop-1 accuracy (%): 62.25 -> 74.52Improved
add decoding-based regularization (prior decoder) (ablation)UniPrior · CODaN classificationTop-1 accuracy (%): 70.26 -> 74.52Improved
add VFM contrastive augmentation (ablation)UniPrior · CODaN classificationTop-1 accuracy (%): 68.89 -> 74.52Improved
use correlation-map alignment instead of naive L1 feature alignment (ablation)UniPrior · CODaN classificationTop-1 accuracy (%): 63.26 -> 74.52Improved
use outlier-robust loss for prior consistency instead of plain L1 (ablation)UniPrior · CODaN classificationTop-1 accuracy (%): 73.63 -> 74.52Improved
add test-time adaptive machine-oriented enhancerUniPrior · CODaN classificationTop-1 accuracy (%): 74.52 -> 75.72Improved
add the TTA enhancer (inference time cost)UniPrior · CODaN classificationinference time (s): 0.005 -> 2.593Hurt
add CLIP-based intermediate loss to enhancer (ablation)UniPrior TTA enhancer · CODaN classificationTop-1 accuracy (%): 74.86 -> 75.72No clear effect
human-oriented low-light enhancement before understandingRefineNet · Nighttime Driving segmentationmIoU (%): 34.3 -> 25.2Hurt
change low-light training distribution for data-requiring adaptationWiiD · CODaN classificationaccuracy drop (%): 1.72% dropHurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy