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

This technical report describes a practical field test on word-imageclassification in a very large collection of more than 300 diverse handwrittenhistorical manuscripts, with 1.6 million unique labeled images and more than 11million images used in testing. Results indicate that several deep-learningtests completely failed (mean accuracy 83%). In the tests with more than 1000output units (lexical words) in one-hot encoding for classification,performance steeply drops to almost zero percent accuracy, even with a modestsize of the pre-final (i.e., penultimate) layer (150 units). A traditionalfeature method (BOVW) displays a consistent performance over numbers of classesand numbers of training examples (mean accuracy 87%). Additional tests usingnearest mean on the output of the pre-final layer of an Inception V3 network,for each book, only yielded mediocre results (mean accuracy 49\%), but was notsensitive to high numbers of classes. Notably, this experiment was onlypossible on the basis of labels that were harvested on the basis of atraditional method which already works starting from a single labeled image perclass. It is expected that the performance of the failed deep learning testscan be repaired, but only on the basis of human handcrafting (sic) of networkarchitecture and hyperparameters. When the failed problematic books are notconsidered, end-to-end CNN training yields about 95% accuracy. This average isdominated by a large subset of Chinese characters, performances for otherscript styles being lower.

A large-scale field test on word-image classification in large historical document collections using a traditional and two deep-learning methods