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

Language-guided human motion synthesis has been a challenging task due to theinherent complexity and diversity of human behaviors. Previous methods facelimitations in generalization to novel actions, often resulting in unrealisticor incoherent motion sequences. In this paper, we propose ATOM (ATomic mOtionModeling) to mitigate this problem, by decomposing actions into atomic actions,and employing a curriculum learning strategy to learn atomic actioncomposition. First, we disentangle complex human motions into a set of atomicactions during learning, and then assemble novel actions using the learnedatomic actions, which offers better adaptability to new actions. Moreover, weintroduce a curriculum learning training strategy that leverages masked motionmodeling with a gradual increase in the mask ratio, and thus facilitates atomicaction assembly. This approach mitigates the overfitting problem commonlyencountered in previous methods while enforcing the model to learn bettermotion representations. We demonstrate the effectiveness of ATOM throughextensive experiments, including text-to-motion and action-to-motion synthesistasks. We further illustrate its superiority in synthesizing plausible andcoherent text-guided human motion sequences.

Language-guided Human Motion Synthesis with Atomic Actions