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

This paper presents an unsupervised transformer-based framework for temporalactivity segmentation which leverages not only frame-level cues but alsosegment-level cues. This is in contrast with previous methods which often relyon frame-level information only. Our approach begins with a frame-levelprediction module which estimates framewise action classes via a transformerencoder. The frame-level prediction module is trained in an unsupervised mannervia temporal optimal transport. To exploit segment-level information, weutilize a segment-level prediction module and a frame-to-segment alignmentmodule. The former includes a transformer decoder for estimating videotranscripts, while the latter matches frame-level features with segment-levelfeatures, yielding permutation-aware segmentation results. Moreover, inspiredby temporal optimal transport, we introduce simple-yet-effective pseudo labelsfor unsupervised training of the above modules. Our experiments on four publicdatasets, i.e., 50 Salads, YouTube Instructions, Breakfast, and DesktopAssembly show that our approach achieves comparable or better performance thanprevious methods in unsupervised activity segmentation.

Permutation-Aware Action Segmentation via Unsupervised Frame-to-Segment Alignment