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

Person text-image matching, also known as text based person search, aims toretrieve images of specific pedestrians using text descriptions. Althoughperson text-image matching has made great research progress, existing methodsstill face two challenges. First, the lack of interpretability of text featuresmakes it challenging to effectively align them with their corresponding imagefeatures. Second, the same pedestrian image often corresponds to multipledifferent text descriptions, and a single text description can correspond tomultiple different images of the same identity. The diversity of textdescriptions and images makes it difficult for a network to extract robustfeatures that match the two modalities. To address these problems, we propose aperson text-image matching method by embedding text-feature interpretabilityand an external attack node. Specifically, we improve the interpretability oftext features by providing them with consistent semantic information with imagefeatures to achieve the alignment of text and describe image region features.Toaddress the challenges posed by the diversity of text and the correspondingperson images, we treat the variation caused by diversity to features as causedby perturbation information and propose a novel adversarial attack and defensemethod to solve it. In the model design, graph convolution is used as the basicframework for feature representation and the adversarial attacks caused by textand image diversity on feature extraction is simulated by implanting anadditional attack node in the graph convolution layer to improve the robustnessof the model against text and image diversity. Extensive experimentsdemonstrate the effectiveness and superiority of text-pedestrian image matchingover existing methods. The source code of the method is published at

Person Text-Image Matching via Text-Feature Interpretability Embedding and External Attack Node Implantation