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

Decoupling spatiotemporal representation refers to decomposing the spatialand temporal features into dimension-independent factors. Although previousRGB-D-based motion recognition methods have achieved promising performancethrough the tightly coupled multi-modal spatiotemporal representation, theystill suffer from (i) optimization difficulty under small data setting due tothe tightly spatiotemporal-entangled modeling;(ii) information redundancy as itusually contains lots of marginal information that is weakly relevant toclassification; and (iii) low interaction between multi-modal spatiotemporalinformation caused by insufficient late fusion. To alleviate these drawbacks,we propose to decouple and recouple spatiotemporal representation forRGB-D-based motion recognition. Specifically, we disentangle the task oflearning spatiotemporal representation into 3 sub-tasks: (1) Learninghigh-quality and dimension independent features through a decoupled spatial andtemporal modeling network. (2) Recoupling the decoupled representation toestablish stronger space-time dependency. (3) Introducing a Cross-modalAdaptive Posterior Fusion (CAPF) mechanism to capture cross-modalspatiotemporal information from RGB-D data. Seamless combination of these noveldesigns forms a robust spatialtemporal representation and achieves betterperformance than state-of-the-art methods on four public motion datasets. Ourcode is available at https://github.com/damo-cv/MotionRGBD.

Decoupling and Recoupling Spatiotemporal Representation for RGB-D-based Motion Recognition