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 Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

With the rapid development of deep learning algorithms, action recognition invideo has achieved many important research results. One issue in actionrecognition, Zero-Shot Action Recognition (ZSAR), has recently attractedconsiderable attention, which classify new categories without any positiveexamples. Another difficulty in action recognition is that untrimmed data mayseriously affect model performance. We propose a composite two-stream frameworkwith a pre-trained model. Our proposed framework includes a classifier branchand a composite feature branch. The graph network model is adopted in each ofthe two branches, which effectively improves the feature extraction andreasoning ability of the framework. In the composite feature branch, a3-channel self-attention models are constructed to weight each frame in thevideo and give more attention to the key frames. Each self-attention modelschannel outputs a set of attention weights to focus on a particular aspect ofthe video, and a set of attention weights corresponds to a one-dimensionalvector. The 3-channel self-attention models can evaluate key frames frommultiple aspects, and the output sets of attention weight vectors form anattention matrix, which effectively enhances the attention of key frames withstrong correlation of action. This model can implement action recognition underzero-shot conditions, and has good recognition performance for untrimmed videodata. Experimental results on relevant data sets confirm the validity of ourmodel.

Action Recognition in Untrimmed Videos with Composite Self-Attention Two-Stream Framework