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

Spatio-Temporal video grounding (STVG) focuses on retrieving thespatio-temporal tube of a specific object depicted by a free-form textualexpression. Existing approaches mainly treat this complicated task as aparallel frame-grounding problem and thus suffer from two types ofinconsistency drawbacks: feature alignment inconsistency and predictioninconsistency. In this paper, we present an end-to-end one-stage framework,termed Spatio-Temporal Consistency-Aware Transformer (STCAT), to alleviatethese issues. Specially, we introduce a novel multi-modal template as theglobal objective to address this task, which explicitly constricts thegrounding region and associates the predictions among all video frames.Moreover, to generate the above template under sufficient video-textualperception, an encoder-decoder architecture is proposed for effective globalcontext modeling. Thanks to these critical designs, STCAT enjoys moreconsistent cross-modal feature alignment and tube prediction without relianceon any pre-trained object detectors. Extensive experiments show that our methodoutperforms previous state-of-the-arts with clear margins on two challengingvideo benchmarks (VidSTG and HC-STVG), illustrating the superiority of theproposed framework to better understanding the association between vision andnatural language. Code is publicly available athttps://github.com/jy0205/STCAT.

Embracing Consistency: A One-Stage Approach for Spatio-Temporal Video Grounding