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

Due to the poor illumination and the difficulty in annotating, nighttimeconditions pose a significant challenge for autonomous vehicle perceptionsystems. Unsupervised domain adaptation (UDA) has been widely applied tosemantic segmentation on such images to adapt models from normal conditions totarget nighttime-condition domains. Self-training (ST) is a paradigm in UDA,where a momentum teacher is utilized for pseudo-label prediction, but aconfirmation bias issue exists. Because the one-directional knowledge transferfrom a single teacher is insufficient to adapt to a large domain shift. Tomitigate this issue, we propose to alleviate domain gap by incrementallyconsidering style influence and illumination change. Therefore, we introduce aone-stage Dual-Teacher Bi-directional Self-training (DTBS) framework for smoothknowledge transfer and feedback. Based on two teacher models, we present anovel pipeline to respectively decouple style and illumination shift. Inaddition, we propose a new Re-weight exponential moving average (EMA) to mergethe knowledge of style and illumination factors, and provide feedback to thestudent model. In this way, our method can be embedded in other UDA methods toenhance their performance. For example, the Cityscapes to ACDC night taskyielded 53.8 mIoU (\%), which corresponds to an improvement of +5\% over theprevious state-of-the-art. The code is available at\url{https://github.com/hf618/DTBS}.

DTBS: Dual-Teacher Bi-directional Self-training for Domain Adaptation in Nighttime Semantic Segmentation