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

Contrastive Language-Image Pretraining (CLIP) has demonstrated impressivezero-shot learning abilities for image understanding, yet limited effort hasbeen made to investigate CLIP for zero-shot video recognition. We introduceOpen-VCLIP, a simple yet effective approach that transforms CLIP into a strongzero-shot video classifier that can recognize unseen actions and events at testtime. Our framework extends CLIP with minimal modifications to modelspatial-temporal relationships in videos, making it a specialized videoclassifier, while striving for generalization. We formally show that trainingan Open-VCLIP is equivalent to continual learning with zero historical data. Toaddress this problem, we propose Interpolated Weight Optimization, whichutilizes the benefit of weight interpolation in both training and test time. Weevaluate our method on three popular and challenging action recognitiondatasets following various zero-shot evaluation protocols and we demonstrateour approach outperforms state-of-the-art methods by clear margins. Inparticular, we achieve 87.9%, 58.3%, 81.1% zero-shot accuracy on UCF, HMDB andKinetics-600 respectively, outperforming state-of-the-art methods by 8.3%, 7.8%and 12.2%. Code is released at https://github.com/wengzejia1/Open-VCLIP.

Open-VCLIP: Transforming CLIP to an Open-vocabulary Video Model via Interpolated Weight Optimization