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/

Zendy Plus

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

Zendy Tools

Zendy Open

The attention-based encoder-decoder framework has recently achievedimpressive results for scene text recognition, and many variants have emergedwith improvements in recognition quality. However, it performs poorly oncontextless texts (e.g., random character sequences) which is unacceptable inmost of real application scenarios. In this paper, we first deeply investigatethe decoding process of the decoder. We empirically find that a representativecharacter-level sequence decoder utilizes not only context information but alsopositional information. Contextual information, which the existing approachesheavily rely on, causes the problem of attention drift. To suppress suchside-effect, we propose a novel position enhancement branch, and dynamicallyfuse its outputs with those of the decoder attention module for scene textrecognition. Specifically, it contains a position aware module to enable theencoder to output feature vectors encoding their own spatial positions, and anattention module to estimate glimpses using the positional clue (i.e., thecurrent decoding time step) only. The dynamic fusion is conducted for morerobust feature via an element-wise gate mechanism. Theoretically, our proposedmethod, dubbed \emph{RobustScanner}, decodes individual characters with dynamicratio between context and positional clues, and utilizes more positional oneswhen the decoding sequences with scarce context, and thus is robust andpractical. Empirically, it has achieved new state-of-the-art results on popularregular and irregular text recognition benchmarks while without muchperformance drop on contextless benchmarks, validating its robustness in bothcontextual and contextless application scenarios.

RobustScanner: Dynamically Enhancing Positional Clues for Robust Text Recognition