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

Recently, Transformer-based methods, which predict polygon points or Beziercurve control points for localizing texts, are popular in scene text detection.However, these methods built upon detection transformer framework might achievesub-optimal training efficiency and performance due to coarse positional querymodeling.In addition, the point label form exploited in previous works impliesthe reading order of humans, which impedes the detection robustness from ourobservation. To address these challenges, this paper proposes a concise DynamicPoint Text DEtection TRansformer network, termed DPText-DETR. In detail,DPText-DETR directly leverages explicit point coordinates to generate positionqueries and dynamically updates them in a progressive way. Moreover, to improvethe spatial inductive bias of non-local self-attention in Transformer, wepresent an Enhanced Factorized Self-Attention module which provides pointqueries within each instance with circular shape guidance. Furthermore, wedesign a simple yet effective positional label form to tackle the side effectof the previous form. To further evaluate the impact of different label formson the detection robustness in real-world scenario, we establish anInverse-Text test set containing 500 manually labeled images. Extensiveexperiments prove the high training efficiency, robustness, andstate-of-the-art performance of our method on popular benchmarks. The code andthe Inverse-Text test set are available athttps://github.com/ymy-k/DPText-DETR.

DPText-DETR: Towards Better Scene Text Detection with Dynamic Points in Transformer