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

This paper presents Scalable Semantic Transfer (SST), a novel trainingparadigm, to explore how to leverage the mutual benefits of the data fromdifferent label domains (i.e. various levels of label granularity) to train apowerful human parsing network. In practice, two common application scenariosare addressed, termed universal parsing and dedicated parsing, where the formeraims to learn homogeneous human representations from multiple label domains andswitch predictions by only using different segmentation heads, and the latteraims to learn a specific domain prediction while distilling the semanticknowledge from other domains. The proposed SST has the following appealingbenefits: (1) it can capably serve as an effective training scheme to embedsemantic associations of human body parts from multiple label domains into thehuman representation learning process; (2) it is an extensible semantictransfer framework without predetermining the overall relations of multiplelabel domains, which allows continuously adding human parsing datasets topromote the training. (3) the relevant modules are only used for auxiliarytraining and can be removed during inference, eliminating the extra reasoningcost. Experimental results demonstrate SST can effectively achieve promisinguniversal human parsing performance as well as impressive improvements comparedto its counterparts on three human parsing benchmarks (i.e.,PASCAL-Person-Part, ATR, and CIHP). Code is available athttps://github.com/yangjie-cv/SST.

Semantic Human Parsing via Scalable Semantic Transfer over Multiple Label Domains