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

Recent adaptive methods for efficient video recognition mostly follow thetwo-stage paradigm of "preview-then-recognition" and have achieved greatsuccess on multiple video benchmarks. However, this two-stage paradigm involvestwo visits of raw frames from coarse-grained to fine-grained during inference(cannot be parallelized), and the captured spatiotemporal features cannot bereused in the second stage (due to varying granularity), being not friendly toefficiency and computation optimization. To this end, inspired by humancognition, we propose a novel recognition paradigm of "View while Moving" forefficient long-untrimmed video recognition. In contrast to the two-stageparadigm, our paradigm only needs to access the raw frame once. The two phasesof coarse-grained sampling and fine-grained recognition are combined intounified spatiotemporal modeling, showing great performance. Moreover, weinvestigate the properties of semantic units in video and propose ahierarchical mechanism to efficiently capture and reason about the unit-leveland video-level temporal semantics in long-untrimmed videos respectively.Extensive experiments on both long-untrimmed and short-trimmed videosdemonstrate that our approach outperforms state-of-the-art methods in terms ofaccuracy as well as efficiency, yielding new efficiency and accuracy trade-offsfor video spatiotemporal modeling.

View while Moving: Efficient Video Recognition in Long-untrimmed Videos