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

Action recognition based on skeleton data has recently witnessed increasingattention and progress. State-of-the-art approaches adopting GraphConvolutional networks (GCNs) can effectively extract features on humanskeletons relying on the pre-defined human topology. Despite associatedprogress, GCN-based methods have difficulties to generalize across domains,especially with different human topological structures. In this context, weintroduce UNIK, a novel skeleton-based action recognition method that is notonly effective to learn spatio-temporal features on human skeleton sequencesbut also able to generalize across datasets. This is achieved by learning anoptimal dependency matrix from the uniform distribution based on a multi-headattention mechanism. Subsequently, to study the cross-domain generalizabilityof skeleton-based action recognition in real-world videos, we re-evaluatestate-of-the-art approaches as well as the proposed UNIK in light of a novelPosetics dataset. This dataset is created from Kinetics-400 videos byestimating, refining and filtering poses. We provide an analysis on how muchperformance improves on smaller benchmark datasets after pre-training onPosetics for the action classification task. Experimental results show that theproposed UNIK, with pre-training on Posetics, generalizes well and outperformsstate-of-the-art when transferred onto four target action classificationdatasets: Toyota Smarthome, Penn Action, NTU-RGB+D 60 and NTU-RGB+D 120.

UNIK: A Unified Framework for Real-world Skeleton-based Action Recognition