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

Creating realistic characters that can react to the users' or anothercharacter's movement can benefit computer graphics, games and virtual realityhugely. However, synthesizing such reactive motions in human-human interactionsis a challenging task due to the many different ways two humans can interact.While there are a number of successful researches in adapting the generativeadversarial network (GAN) in synthesizing single human actions, there are veryfew on modelling human-human interactions. In this paper, we propose asemi-supervised GAN system that synthesizes the reactive motion of a charactergiven the active motion from another character. Our key insights are two-fold.First, to effectively encode the complicated spatial-temporal information of ahuman motion, we empower the generator with a part-based long short-term memory(LSTM) module, such that the temporal movement of different limbs can beeffectively modelled. We further include an attention module such that thetemporal significance of the interaction can be learned, which enhances thetemporal alignment of the active-reactive motion pair. Second, as the reactivemotion of different types of interactions can be significantly different, weintroduce a discriminator that not only tells if the generated movement isrealistic or not, but also tells the class label of the interaction. Thisallows the use of such labels in supervising the training of the generator. Weexperiment with the SBU and the HHOI datasets. The high quality of thesynthetic motion demonstrates the effective design of our generator, and thediscriminability of the synthesis also demonstrates the strength of ourdiscriminator.

GAN-based Reactive Motion Synthesis with Class-aware Discriminators for Human-human Interaction