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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Vision-language models (VLMs) pre-trained on large-scale image-text pairshave demonstrated impressive transferability on various visual tasks.Transferring knowledge from such powerful VLMs is a promising direction forbuilding effective video recognition models. However, current exploration inthis field is still limited. We believe that the greatest value of pre-trainedVLMs lies in building a bridge between visual and textual domains. In thispaper, we propose a novel framework called BIKE, which utilizes the cross-modalbridge to explore bidirectional knowledge: i) We introduce the Video AttributeAssociation mechanism, which leverages the Video-to-Text knowledge to generatetextual auxiliary attributes for complementing video recognition. ii) We alsopresent a Temporal Concept Spotting mechanism that uses the Text-to-Videoexpertise to capture temporal saliency in a parameter-free manner, leading toenhanced video representation. Extensive studies on six popular video datasets,including Kinetics-400 & 600, UCF-101, HMDB-51, ActivityNet and Charades, showthat our method achieves state-of-the-art performance in various recognitionscenarios, such as general, zero-shot, and few-shot video recognition. Our bestmodel achieves a state-of-the-art accuracy of 88.6% on the challengingKinetics-400 using the released CLIP model. The code is available athttps://github.com/whwu95/BIKE .

Bidirectional Cross-Modal Knowledge Exploration for Video Recognition with Pre-trained Vision-Language Models