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/

Zendy Plus

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

Zendy Tools

Zendy Open

Dense video captioning aims to generate corresponding text descriptions for aseries of events in the untrimmed video, which can be divided into twosub-tasks, event detection and event captioning. Unlike previous works thattackle the two sub-tasks separately, recent works have focused on enhancing theinter-task association between the two sub-tasks. However, designing inter-taskinteractions for event detection and captioning is not trivial due to the largedifferences in their task specific solutions. Besides, previous event detectionmethods normally ignore temporal dependencies between events, leading to eventredundancy or inconsistency problems. To tackle above the two defects, in thispaper, we define event detection as a sequence generation task and propose aunified pre-training and fine-tuning framework to naturally enhance theinter-task association between event detection and captioning. Since the modelpredicts each event with previous events as context, the inter-dependencybetween events is fully exploited and thus our model can detect more diverseand consistent events in the video. Experiments on the ActivityNet dataset showthat our model outperforms the state-of-the-art methods, and can be furtherboosted when pre-trained on extra large-scale video-text data. Code isavailable at \url{https://github.com/QiQAng/UEDVC}.

Unifying Event Detection and Captioning as Sequence Generation via Pre-Training