Neural Network lernt Handschriften



Tom Haines vom University College London hat ein Neural Network entwickelt, das Handschriften lernen kann, hier sein Blog-Eintrag zum Paper.

Nettes Gimmick im algorithmisch generierten Sherlock-Holmes-Zitat in der Handschrift von Sir Arthur Conan Doyle oben: Er hat diesen Satz („Elementary my dear Watson“) nie geschrieben, der stammt aus den 1929er Film The Return of Sherlock Holmes.

Vor drei Jahren gab's bereits eine ähnliche Arbeit von Alex Graves aus Kanada (damals gab's auch einen netten Online-Generator dazu). Die „Hand“-Schriften aus Graves Algorithmus waren noch relativ wackelig, Haines' AI scheint da merklich Fortschritte im Schriftbild gemacht zu habe, wobei die Grundlinie nach wie vor teilweise extrem springt. Menschen schreiben zwar auch nie in einer hundertprozentig graden Linie, aber sie springen nicht innerhalb der Worte ein paar Millimeter hoch.

There are many scenarios where we wish to imitate a specific author's pen-on-paper handwriting style. Rendering new text in someone's handwriting is difficult because natural handwriting is highly variable, yet follows both intentional and involuntary structure that makes a person's style self-consistent.

We present an algorithm that renders a desired input string in an author's handwriting. An annotated sample of the author's handwriting is required; the system is flexible enough that historical documents can usually be used with only a little extra effort. Experiments show that our glyph-centric approach, with learned parameters for spacing, line thickness, and pressure, produces novel images of handwriting that look hand-made to casual observers, even when printed on paper.


Mehr bei der BBC: Can a computer copy your handwriting?

They have created an algorithm that can take a sample of handwritten text, examine its qualities, and then write any text in the same style.

There are already typefaces in word processing programs that produce text in a fairly uniform handwritten style. But what Tom Haines and his fellow UCL researchers have done is create software that they claim reproduces the messy details of any individual writer's hand.

They call their system My Text In Your Handwriting and have tried it out on samples of handwritten text from historical figures such as Abraham Lincoln and the creator of Sherlock Holmes Sir Arthur Conan Doyle.