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

In this paper, we introduce the concept of learning latent super-events fromactivity videos, and present how it benefits activity detection in continuousvideos. We define a super-event as a set of multiple events occurring togetherin videos with a particular temporal organization; it is the opposite conceptof sub-events. Real-world videos contain multiple activities and are rarelysegmented (e.g., surveillance videos), and learning latent super-events allowsthe model to capture how the events are temporally related in videos. We designtemporal structure filters that enable the model to focus on particularsub-intervals of the videos, and use them together with a soft attentionmechanism to learn representations of latent super-events. Super-eventrepresentations are combined with per-frame or per-segment CNNs to provideframe-level annotations. Our approach is designed to be fully differentiable,enabling end-to-end learning of latent super-event representations jointly withthe activity detector using them. Our experiments with multiple public videodatasets confirm that the proposed concept of latent super-event learningsignificantly benefits activity detection, advancing the state-of-the-arts.

Learning Latent Super-Events to Detect Multiple Activities in Videos