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

Human action recognition refers to automatic recognizing human actions from avideo clip. In reality, there often exist multiple human actions in a videostream. Such a video stream is often weakly-annotated with a set of relevanthuman action labels at a global level rather than assigning each label to aspecific video episode corresponding to a single action, which leads to amulti-label learning problem. Furthermore, there are many meaningful humanactions in reality but it would be extremely difficult to collect/annotatevideo clips regarding all of various human actions, which leads to a zero-shotlearning scenario. To the best of our knowledge, there is no work that hasaddressed all the above issues together in human action recognition. In thispaper, we formulate a real-world human action recognition task as a multi-labelzero-shot learning problem and propose a framework to tackle this problem in aholistic way. Our framework holistically tackles the issue of unknown temporalboundaries between different actions for multi-label learning and exploits theside information regarding the semantic relationship between different humanactions for knowledge transfer. Consequently, our framework leads to a jointlatent ranking embedding for multi-label zero-shot human action recognition. Anovel neural architecture of two component models and an alternate learningalgorithm are proposed to carry out the joint latent ranking embeddinglearning. Thus, multi-label zero-shot recognition is done by measuringrelatedness scores of action labels to a test video clip in the joint latentvisual and semantic embedding spaces. We evaluate our framework with differentsettings, including a novel data split scheme designed especially forevaluating multi-label zero-shot learning, on two datasets: Breakfast andCharades. The experimental results demonstrate the effectiveness of ourframework.

Multi-Label Zero-Shot Human Action Recognition via Joint Latent Ranking Embedding