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

The divergence between labeled training data and unlabeled testing data is asignificant challenge for recent deep learning models. Unsupervised domainadaptation (UDA) attempts to solve such problem. Recent works show thatself-training is a powerful approach to UDA. However, existing methods havedifficulty in balancing the scalability and performance. In this paper, wepropose a hard-aware instance adaptive self-training framework for UDA on thetask of semantic segmentation. To effectively improve the quality and diversityof pseudo-labels, we develop a novel pseudo-label generation strategy with aninstance adaptive selector. We further enrich the hard class pseudo-labels withinter-image information through a skillfully designed hard-aware pseudo-labelaugmentation. Besides, we propose the region-adaptive regularization to smooththe pseudo-label region and sharpen the non-pseudo-label region. For thenon-pseudo-label region, consistency constraint is also constructed tointroduce stronger supervision signals during model optimization. Our method isso concise and efficient that it is easy to be generalized to other UDAmethods. Experiments on GTA5 to Cityscapes, SYNTHIA to Cityscapes, andCityscapes to Oxford RobotCar demonstrate the superior performance of ourapproach compared with the state-of-the-art methods.

Hard-aware Instance Adaptive Self-training for Unsupervised Cross-domain Semantic Segmentation