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

Sequential video understanding, as an emerging video understanding task, hasdriven lots of researchers' attention because of its goal-oriented nature. Thispaper studies weakly supervised sequential video understanding where theaccurate time-stamp level text-video alignment is not provided. We solve thistask by borrowing ideas from CLIP. Specifically, we use a transformer toaggregate frame-level features for video representation and use a pre-trainedtext encoder to encode the texts corresponding to each action and the wholevideo, respectively. To model the correspondence between text and video, wepropose a multiple granularity loss, where the video-paragraph contrastive lossenforces matching between the whole video and the complete script, and afine-grained frame-sentence contrastive loss enforces the matching between eachaction and its description. As the frame-sentence correspondence is notavailable, we propose to use the fact that video actions happen sequentially inthe temporal domain to generate pseudo frame-sentence correspondence andsupervise the network training with the pseudo labels. Extensive experiments onvideo sequence verification and text-to-video matching show that our methodoutperforms baselines by a large margin, which validates the effectiveness ofour proposed approach. Code is available at https://github.com/svip-lab/WeakSVR

Weakly Supervised Video Representation Learning with Unaligned Text for Sequential Videos