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

Unsupervised skeleton based action recognition has achieved remarkableprogress recently. Existing unsupervised learning methods suffer from severeoverfitting problem, and thus small networks are used, significantly reducingthe representation capability. To address this problem, the overfittingmechanism behind the unsupervised learning for skeleton based actionrecognition is first investigated. It is observed that the skeleton is alreadya relatively high-level and low-dimension feature, but not in the same manifoldas the features for action recognition. Simply applying the existingunsupervised learning method may tend to produce features that discriminate thedifferent samples instead of action classes, resulting in the overfittingproblem. To solve this problem, this paper presents an Unsupervisedspatial-temporal Feature Enrichment and Fidelity Preservation framework(U-FEFP) to generate rich distributed features that contain all the informationof the skeleton sequence. A spatial-temporal feature transformation subnetworkis developed using spatial-temporal graph convolutional network and graphconvolutional gate recurrent unit network as the basic feature extractionnetwork. The unsupervised Bootstrap Your Own Latent based learning is used togenerate rich distributed features and the unsupervised pretext task basedlearning is used to preserve the information of the skeleton sequence. The twounsupervised learning ways are collaborated as U-FEFP to produce robust anddiscriminative representations. Experimental results on three widely usedbenchmarks, namely NTU-RGB+D-60, NTU-RGB+D-120 and PKU-MMD dataset, demonstratethat the proposed U-FEFP achieves the best performance compared with thestate-of-the-art unsupervised learning methods. t-SNE illustrations furthervalidate that U-FEFP can learn more discriminative features for unsupervisedskeleton based action recognition.

Unsupervised Spatial-Temporal Feature Enrichment and Fidelity Preservation Network for Skeleton based Action Recognition