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

Scene text detection and recognition have been well explored in the past fewyears. Despite the progress, efficient and accurate end-to-end spotting ofarbitrarily-shaped text remains challenging. In this work, we propose anend-to-end text spotting framework, termed PAN++, which can efficiently detectand recognize text of arbitrary shapes in natural scenes. PAN++ is based on thekernel representation that reformulates a text line as a text kernel (centralregion) surrounded by peripheral pixels. By systematically comparing withexisting scene text representations, we show that our kernel representation cannot only describe arbitrarily-shaped text but also well distinguish adjacenttext. Moreover, as a pixel-based representation, the kernel representation canbe predicted by a single fully convolutional network, which is very friendly toreal-time applications. Taking the advantages of the kernel representation, wedesign a series of components as follows: 1) a computationally efficientfeature enhancement network composed of stacked Feature Pyramid EnhancementModules (FPEMs); 2) a lightweight detection head cooperating with PixelAggregation (PA); and 3) an efficient attention-based recognition head withMasked RoI. Benefiting from the kernel representation and the tailoredcomponents, our method achieves high inference speed while maintainingcompetitive accuracy. Extensive experiments show the superiority of our method.For example, the proposed PAN++ achieves an end-to-end text spotting F-measureof 64.9 at 29.2 FPS on the Total-Text dataset, which significantly outperformsthe previous best method. Code will be available at: https://git.io/PAN.

PAN++: Towards Efficient and Accurate End-to-End Spotting of Arbitrarily-Shaped Text