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

In this paper, we consider the problem of temporal action localization underlow-shot (zero-shot & few-shot) scenario, with the goal of detecting andclassifying the action instances from arbitrary categories within someuntrimmed videos, even not seen at training time. We adopt a Transformer-basedtwo-stage action localization architecture with class-agnostic action proposal,followed by open-vocabulary classification. We make the followingcontributions. First, to compensate image-text foundation models with temporalmotions, we improve category-agnostic action proposal by explicitly aligningembeddings of optical flows, RGB and texts, which has largely been ignored inexisting low-shot methods. Second, to improve open-vocabulary actionclassification, we construct classifiers with strong discriminative power,i.e., avoid lexical ambiguities. To be specific, we propose to prompt thepre-trained CLIP text encoder either with detailed action descriptions(acquired from large-scale language models), or visually-conditionedinstance-specific prompt vectors. Third, we conduct thorough experiments andablation studies on THUMOS14 and ActivityNet1.3, demonstrating the superiorperformance of our proposed model, outperforming existing state-of-the-artapproaches by one significant margin.

Multi-modal Prompting for Low-Shot Temporal Action Localization