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

Domain shift widely exists in the visual world, while modern deep neuralnetworks commonly suffer from severe performance degradation under domain shiftdue to the poor generalization ability, which limits the real-worldapplications. The domain shift mainly lies in the limited source environmentalvariations and the large distribution gap between source and unseen targetdata. To this end, we propose a unified framework, Style-HAllucinated DualconsistEncy learning (SHADE), to handle such domain shift in various visualtasks. Specifically, SHADE is constructed based on two consistency constraints,Style Consistency (SC) and Retrospection Consistency (RC). SC enriches thesource situations and encourages the model to learn consistent representationacross style-diversified samples. RC leverages general visual knowledge toprevent the model from overfitting to source data and thus largely keeps therepresentation consistent between the source and general visual models.Furthermore, we present a novel style hallucination module (SHM) to generatestyle-diversified samples that are essential to consistency learning. SHMselects basis styles from the source distribution, enabling the model todynamically generate diverse and realistic samples during training. Extensiveexperiments demonstrate that our versatile SHADE can significantly enhance thegeneralization in various visual recognition tasks, including imageclassification, semantic segmentation and object detection, with differentmodels, i.e., ConvNets and Transformer.

Style-Hallucinated Dual Consistency Learning: A Unified Framework for Visual Domain Generalization