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

As the most essential property in a video, motion information is critical toa robust and generalized video representation. To inject motion dynamics,recent works have adopted frame difference as the source of motion informationin video contrastive learning, considering the trade-off between quality andcost. However, existing works align motion features at the instance level,which suffers from spatial and temporal weak alignment across modalities. Inthis paper, we present a \textbf{Fi}ne-grained \textbf{M}otion\textbf{A}lignment (FIMA) framework, capable of introducing well-aligned andsignificant motion information. Specifically, we first develop a densecontrastive learning framework in the spatiotemporal domain to generatepixel-level motion supervision. Then, we design a motion decoder and aforeground sampling strategy to eliminate the weak alignments in terms of timeand space. Moreover, a frame-level motion contrastive loss is presented toimprove the temporal diversity of the motion features. Extensive experimentsdemonstrate that the representations learned by FIMA possess greatmotion-awareness capabilities and achieve state-of-the-art or competitiveresults on downstream tasks across UCF101, HMDB51, and Diving48 datasets. Codeis available at \url{https://github.com/ZMHH-H/FIMA}.

Fine-Grained Spatiotemporal Motion Alignment for Contrastive Video Representation Learning