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

We aim for zero-shot localization and classification of human actions invideo. Where traditional approaches rely on global attribute or objectclassification scores for their zero-shot knowledge transfer, our maincontribution is a spatial-aware object embedding. To arrive at spatialawareness, we build our embedding on top of freely available actor and objectdetectors. Relevance of objects is determined in a word embedding space andfurther enforced with estimated spatial preferences. Besides local objectawareness, we also embed global object awareness into our embedding to maximizeactor and object interaction. Finally, we exploit the object positions andsizes in the spatial-aware embedding to demonstrate a new spatio-temporalaction retrieval scenario with composite queries. Action localization andclassification experiments on four contemporary action video datasets supportour proposal. Apart from state-of-the-art results in the zero-shot localizationand classification settings, our spatial-aware embedding is even competitivewith recent supervised action localization alternatives.

Spatial-Aware Object Embeddings for Zero-Shot Localization and Classification of Actions