Re: Re: scribble update

On Mon, Sep 11, 2000 at 01:08:40PM -0700, Jim Gettys wrote:
> BTW, in my wandering I came across:
>
> http://www.itl.nist.gov/iaui/vip/databases/defs/nist_ocr.html
>
> Looks interesting: may have been what the guy who was in our booth at Usenix
> was talking about...
> - Jim

I actually have the CD of this (as well as Special Database 1: Binary
Images of Printed Digits, Alphas,and Text), but haven't done anything with
it. The software is designed for doing batch OCR on scans of handprinted,
preprinted forms with specific test cases (can you say census?)

As a consequence, it doesn't have the stroke information available, just
the pixels, and does a lot of work to remove the 'known' part of the image
(the preprinted form) to retrieve the actually interesting pixels.

There's probably interesting code in there, lots of matrix math doing
feature recognition, etc., but nothing plug and play, for a stroke
oriented, realtime recognition environment. If someone's interested,
I could throw the CD up somewhere for web access so you could poke
around. It's available free of charge from NIST, as the URL you went to
points out.

This is actually an area of software I'm personally interested in. I've
got three devices now (soon 4, once my iPaq shows up) that can capture
stroke input: a Palm Pilot, a CrossPad (that one's sort of interesting,
since it's stored strokes: http://www.crosspad.com/cross/crosspads.html)
and an eBeam whiteboard add on, which captures (and shares) my physical
whiteboard (it's at www.e-beam.com, not ebeam: that's electron beam
lithography).

A unified API for accessing stroke info, and apps that can deal with
stored strokes would be very interesting.

Ross

-- 
Ross J. Reedstrom, Ph.D., <reedstrm@rice.edu> 
NSBRI Research Scientist/Programmer
Computer and Information Technology Institute
Rice University, 6100 S. Main St.,  Houston, TX 77005