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

Weakly supervised temporal action localization (WSTAL) aims to localizeactions in untrimmed videos using video-level labels. Despite recent advances,existing approaches mainly follow a localization-by-classification pipeline,generally processing each segment individually, thereby exploiting only limitedcontextual information. As a result, the model will lack a comprehensiveunderstanding (e.g. appearance and temporal structure) of various actionpatterns, leading to ambiguity in classification learning and temporallocalization. Our work addresses this from a novel perspective, by exploringand exploiting the cross-video contextual knowledge within the dataset torecover the dataset-level semantic structure of action instances via weaklabels only, thereby indirectly improving the holistic understanding offine-grained action patterns and alleviating the aforementioned ambiguities.Specifically, an end-to-end framework is proposed, including a RobustMemory-Guided Contrastive Learning (RMGCL) module and a Global KnowledgeSummarization and Aggregation (GKSA) module. First, the RMGCL module exploresthe contrast and consistency of cross-video action features, assisting inlearning more structured and compact embedding space, thus reducing ambiguityin classification learning. Further, the GKSA module is used to efficientlysummarize and propagate the cross-video representative action knowledge in alearnable manner to promote holistic action patterns understanding, which inturn allows the generation of high-confidence pseudo-labels for self-learning,thus alleviating ambiguity in temporal localization. Extensive experiments onTHUMOS14, ActivityNet1.3, and FineAction demonstrate that our methodoutperforms the state-of-the-art methods, and can be easily plugged into otherWSTAL methods.

Cross-Video Contextual Knowledge Exploration and Exploitation for Ambiguity Reduction in Weakly Supervised Temporal Action Localization