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

Temporal action detection (TAD) involves the localization and classificationof action instances within untrimmed videos. While standard TAD follows fullysupervised learning with closed-set setting on large training data, recentzero-shot TAD methods showcase the promising open-set setting by leveraginglarge-scale contrastive visual-language (ViL) pretrained models. However,existing zero-shot TAD methods have limitations on how to properly constructthe strong relationship between two interdependent tasks of localization andclassification and adapt ViL model to video understanding. In this work, wepresent ZEETAD, featuring two modules: dual-localization and zero-shot proposalclassification. The former is a Transformer-based module that detects actionevents while selectively collecting crucial semantic embeddings for laterrecognition. The latter one, CLIP-based module, generates semantic embeddingsfrom text and frame inputs for each temporal unit. Additionally, we enhancediscriminative capability on unseen classes by minimally updating the frozenCLIP encoder with lightweight adapters. Extensive experiments on THUMOS14 andActivityNet-1.3 datasets demonstrate our approach's superior performance inzero-shot TAD and effective knowledge transfer from ViL models to unseen actioncategories.

ZEETAD: Adapting Pretrained Vision-Language Model for Zero-Shot End-to-End Temporal Action Detection