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

RGB-Infrared (IR) person re-identification aims to retrieveperson-of-interest from heterogeneous cameras, easily suffering from largeimage modality discrepancy caused by different sensing wavelength ranges.Existing work usually minimizes such discrepancy by aligning domaindistribution of global features, while neglecting the intra-modality structuralrelations between semantic parts. This could result in the network overlyfocusing on local cues, without considering long-range body part dependencies,leading to meaningless region representations. In this paper, we propose agraph-enabled distribution matching solution, dubbed Geometry-GuidedDual-Alignment (G2DA) learning, for RGB-IR ReID. It can jointly encourage thecross-modal consistency between part semantics and structural relations forfine-grained modality alignment by solving a graph matching task within amulti-scale skeleton graph that embeds human topology information.Specifically, we propose to build a semantic-aligned complete graph into whichall cross-modality images can be mapped via a pose-adaptive graph constructionmechanism. This graph represents extracted whole-part features by nodes andexpresses the node-wise similarities with associated edges. To achieve thegraph-based dual-alignment learning, an Optimal Transport (OT) based structuredmetric is further introduced to simultaneously measure point-wise relations andgroup-wise structural similarities across modalities. By minimizing the cost ofan inter-modality transport plan, G2DA can learn a consistent anddiscriminative feature subspace for cross-modality image retrieval.Furthermore, we advance a Message Fusion Attention (MFA) mechanism toadaptively reweight the information flow of semantic propagation, effectivelystrengthening the discriminability of extracted semantic features.

G2DA: Geometry-Guided Dual-Alignment Learning for RGB-Infrared Person Re-Identification