English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Nighttime semantic segmentation is essential for various applications, e.g.,autonomous driving, which often faces challenges due to poor illumination andthe lack of well-annotated datasets. Unsupervised domain adaptation (UDA) hasshown potential for addressing the challenges and achieved remarkable resultsfor nighttime semantic segmentation. However, existing methods still facelimitations in 1) their reliance on style transfer or relighting models, whichstruggle to generalize to complex nighttime environments, and 2) theirignorance of dynamic and small objects like vehicles and traffic signs, whichare difficult to be directly learned from other domains. This paper proposes anovel UDA method that refines both label and feature levels for dynamic andsmall objects for nighttime semantic segmentation. First, we propose a dynamicand small object refinement module to complement the knowledge of dynamic andsmall objects from the source domain to target nighttime domain. These dynamicand small objects are normally context-inconsistent in under-exposedconditions. Then, we design a feature prototype alignment module to reduce thedomain gap by deploying contrastive learning between features and prototypes ofthe same class from different domains, while re-weighting the categories ofdynamic and small objects. Extensive experiments on four benchmark datasetsdemonstrate that our method outperforms prior arts by a large margin fornighttime segmentation. Project page: https://rorisis.github.io/DSRNSS/.

Towards Dynamic and Small Objects Refinement for Unsupervised Domain Adaptative Nighttime Semantic Segmentation