What is Optical Glyph Recognition (OGR)?
fabrizio.borgia at GMAIL.COM
Wed Nov 6 09:19:42 UTC 2013
Thanks Maria, no worries, anytime will be OK!
Since we are focusing on handwritten SW texts, we did not run a relevant
number of tests on computer printed material. Despite this, the
computer-printed material has been very useful in the first phase of the
development, since we used it to test our geometric recognition routines
(detection of rectangles, squares, circles, etc.) BEFORE dealing with
handwritten material, which proved much harder to recognize, and brought up
some interesting issues (just think about recognizing an incomplete
rectangle, or a badly drawn circle) which are now, in large part, solved.
Anyway, it will be definitely possible to print a Wikipedia ASL page, scan
it, and feed it to our OGR Engine!
As for formal SignWriting, this perspective is very interesting, but I'm
afraid I don't know much about it (SignWriting has so many projects
underway!) I'll be documenting and working on it as soon as I can!
Thanks for asking about the "digital format", I have not completely clear.
Yes, I need handwritten text, written in columns.
With "digital format" I was actually meaning "digital image format" so any
image or pdf format will do (.png .jpg .pdf etc...).
The great deal of documents sitting in your garage, is REALLY one of the
main reason for us to renew our efforts on OGR.
Despite this, it's not my intention to take away much time from you. Just
a few scans will do!
We are working on the modern (ISWA2010) version of SignWriting, so, even if
it would be indeed very interesting to see SW variations from 1974 to
today, I'm afraid we will need just the latest material :(
About the ASL Wikipedia project:
I know the project, and I've written about it in my Ph.D. thesis, since I
think it is one of the most interesting SW projects at the moment. I really
had FUN reading the Abraham Lincoln page, and recognizing the few things I
know of ASL while I was reproduding the signs I was reading (large use of
initialized signs, in particular)! We'd really be excited to contribute, as
soon as the workload of OGR will get lighter!
*Thank you so much, everybody!*
2013/11/5 Valerie Sutton <signwriting at mac.com>
> SignWriting List
> November 5, 2013
> Thank you, Fabrizio…
> Do you want only “handwritten” SignWriting documents in vertical columns?
> I had gotten the impression that your project wanted only “writing by hand”
> kind of documents?
> But then you mention in any digital format….so could that be computer
> written texts as well?
> I have a lot of handwritten documents from our past, in my garage (can you
> believe?), that are sitting in notebooks…mostly written with ink pens in
> stick figures from the early 1980’s for the SignWriter Newspaper project.
> None of it is computer written because we did not have computer programs at
> that time - so it is all written by hand with ink pens and most of it looks
> better than the average handwriting because the purpose of most of the
> documents was to be published as articles in the SignWriter Newspaper. Most
> of it is stick figures written receptively and horizontally from left to
> right -
> But I also have other documents of individuals around the world writing by
> hand too…but not all of it is in vertical columns - some in stick figures
> and some not -
> So my point is, do you want to see the enormous variation of handwritten
> documents, from 1974-today? or do you want to see computer-written texts
> too, like the ASL Wikipedia? Steve mentioned that it a previous message and
> there are some cool things happening with the ASL Wkipedia project and
> character recognition for that project would indeed be wonderful -
> Regarding the ASL Wikipedia project….
> Steve is in the process of moving our ASL Wikipedia site from a test area
> on Wikimedia Labs to the Wikimedia Incubator server. The goal of course is
> to someday become an “official” Wikipedia, but there is more software
> development to complete before that can happen. We are excited to be moving
> over to the Incubator web site. At the moment, the site can be best viewed
> if you use Chrome or Safari web browsers….Below is a link to a page that
> lists the titles of some of the documents that have been moved to the
> Incubator…and as you can see the entire site, including the Menus and the
> Titles for the articles, are written in ASL. In time, there will be a way
> to toggle back and forth between languages so the Menus can change
> languages later, if you cannot read a Menu item in ASL...
> Using Chrome or Safari, go to Titles and click on a title in SignWriting
> to view an article in ASL:
> So give me some guidelines - should I scan in pages of handwritten
> documents from the 1980s? Or do you want more modern documents the way the
> world is writing right now? and is it only hand-written documents that you
> Val ;-)
> On Nov 4, 2013, at 6:57 AM, Stephen E Slevinski Jr <slevin at signpuddle.net>
> Hi Fabrizio,
> Thanks for keeping us posted. What results have you had from computer
> printed material?
> Would it be possible to print a page from the American Sign Language
> Wikipedia (SVG) and use Optical Glyph Recognition to identify the symbols
> Simple math would identify their 2-dimensional placement. It sounds like
> it would be possible to create Formal SignWriting strings from a scan. The
> strings wouldn't be exactly the same as the input, but searching in
> SignWriting doesn't require exact string matches. This is very exciting.
> On Nov 4, 2013, at 4:50 AM, Fabrizio Borgia <fabrizio.borgia at GMAIL.COM>
> Dear Valerie,
> we are absolutely in no rush, since we are still developing the OGR
> engine, and it is still far from complete.
> As soon as the new material arrives, I'll be happy to run some tests and
> show you the result!
> 2013/10/31 Valerie Sutton <signwriting at mac.com>
>> SignWriting List
>> October 31, 2013
>> Hello All, Fabrizio, and the OGR Team in Toulouse, France ;-)
>> Happy to meet you all -
>> Thank you for this important email, and I am very happy to know about
>> your project. I want you to know that I have not forgotten you…but have
>> been waiting for a quiet moment to be able to concentrate and to provide
>> you with all the feedback and materials you need…So you are in my thoughts,
>> and a real answer is coming - All is well here… Just have to take one step
>> at a time… I look forward to more communication this weekend - Val ;-)
>> On Oct 29, 2013, at 9:13 AM, Fabrizio Borgia <fabrizio.borgia at GMAIL.COM>
>> Dear All,
>> I'd like to take a moment to explain the work of my team (Université
>> Toulouse III - Paul Sabatier, Computer Science Department - "Sapienza"
>> University of Rome) in the latest few years.
>> We are working at a *Optical Glyph Recogition* (OGR) engine for
>> Well, OGR sounds very much like OCR (which is Optical Character
>> Recognition) because they both share the same goal: *the conversion of
>> images of text into characters*. In the OGR case, the characters are the
>> Apart from the goal, OGR and OCR have very little in common. The design
>> and the implementation of our application is somewhat different from the
>> standard OCR design, because we are facing different problems, and
>> different datasets.
>> We know that there's a huge amount of handwritten SW corpora, scattered
>> around the world, and we believe that they carry an invaluable knowledge,
>> both for the Deaf Community, and for the Research Community. Our goal is to
>> make them digital, to *empower their diffusion*, and to *ensure a
>> long-lasting existence* for them.
>> After a very, very thorough design phase, we are developing the OGR
>> application (and, in confidence, we are about halfway to go) and *soon
>> we will release a footage showing our software in action*.
>> To make our OGR stronger, we need to test it with diffrent datasets (SW
>> texts, pages, etc.) we need your help, dear SW list!
>> *We need any kind of handwritten SW texts, in any digital image format. *
>> *Optionally, we'd like them to be:*
>> - *written in columns*
>> *whithout any segmentation (annotation lines, squares, etc.) from one
>> sign to the following one *
>> Anyone can send us some material (even spare pages will DEFINITLEY do) or
>> point us in the right direction?
>> Thanks in advance for your time, from Rome!
>> Fabrizio Borgia, M.D.
>> Ph.D. Student
>> Sapienza - Università di Roma
>> Université Toulouse III - Paul Sabatier
>> O&M SW Engineering Intern
>> SERCO S.p.A.
>> fabrizio.borgia at gmail.com
>> fabrizio.borgia at uniroma1.it
>> fabrizio.borgia at serco.com
> SIGNWRITING LIST INFORMATION
> Valerie Sutton SignWriting List moderator sutton at signwriting.org
> Post Messages to the SignWriting List: sw-l at listserv.valenciacollege.edu
> SignWriting List Archives & Home Page
> Join, Leave or Change How You Receive SW List Messages
Fabrizio Borgia, M.D.
Sapienza - Università di Roma
Université Toulouse III - Paul Sabatier
O&M SW Engineering Intern
fabrizio.borgia at gmail.com
fabrizio.borgia at uniroma1.it
fabrizio.borgia at serco.com
SIGNWRITING LIST INFORMATION
SignWriting List moderator
sutton at signwriting.org
Post Messages to the SignWriting List:
sw-l at listserv.valenciacollege.edu
SignWriting List Archives & Home Page
Join, Leave or Change How You Receive SW List Messages
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Sw-l