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

The Transformer-based encoder-decoder framework is becoming popular in scenetext recognition, largely because it naturally integrates recognition cluesfrom both visual and semantic domains. However, recent studies show that thetwo kinds of clues are not always well registered and therefore, feature andcharacter might be misaligned in difficult text (e.g., with a rare shape). As aresult, constraints such as character position are introduced to alleviate thisproblem. Despite certain success, visual and semantic are still separatelymodeled and they are merely loosely associated. In this paper, we propose anovel module called Multi-Domain Character Distance Perception (MDCDP) toestablish a visually and semantically related position embedding. MDCDP usesthe position embedding to query both visual and semantic features following thecross-attention mechanism. The two kinds of clues are fused into the positionbranch, generating a content-aware embedding that well perceives characterspacing and orientation variants, character semantic affinities, and cluestying the two kinds of information. They are summarized as the multi-domaincharacter distance. We develop CDistNet that stacks multiple MDCDPs to guide agradually precise distance modeling. Thus, the feature-character alignment iswell built even various recognition difficulties are presented. We verifyCDistNet on ten challenging public datasets and two series of augmenteddatasets created by ourselves. The experiments demonstrate that CDistNetperforms highly competitively. It not only ranks top-tier in standardbenchmarks, but also outperforms recent popular methods by obvious margins onreal and augmented datasets presenting severe text deformation, poor linguisticsupport, and rare character layouts. Code is available athttps://github.com/simplify23/CDistNet.

CDistNet: Perceiving Multi-Domain Character Distance for Robust Text Recognition