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

Contrastive learning has been successfully leveraged to learn actionrepresentations for addressing the problem of semi-supervised skeleton-basedaction recognition. However, most contrastive learning-based methods onlycontrast global features mixing spatiotemporal information, which confuses thespatial- and temporal-specific information reflecting different semantic at theframe level and joint level. Thus, we propose a novel SpatiotemporalDecouple-and-Squeeze Contrastive Learning (SDS-CL) framework to comprehensivelylearn more abundant representations of skeleton-based actions by jointlycontrasting spatial-squeezing features, temporal-squeezing features, and globalfeatures. In SDS-CL, we design a new Spatiotemporal-decoupling Intra-InterAttention (SIIA) mechanism to obtain the spatiotemporal-decoupling attentivefeatures for capturing spatiotemporal specific information by calculatingspatial- and temporal-decoupling intra-attention maps among joint/motionfeatures, as well as spatial- and temporal-decoupling inter-attention mapsbetween joint and motion features. Moreover, we present a new Spatial-squeezingTemporal-contrasting Loss (STL), a new Temporal-squeezing Spatial-contrastingLoss (TSL), and the Global-contrasting Loss (GL) to contrast thespatial-squeezing joint and motion features at the frame level,temporal-squeezing joint and motion features at the joint level, as well asglobal joint and motion features at the skeleton level. Extensive experimentalresults on four public datasets show that the proposed SDS-CL achievesperformance gains compared with other competitive methods.

Spatiotemporal Decouple-and-Squeeze Contrastive Learning for Semi-Supervised Skeleton-based Action Recognition