Klarinet Archive - Posting 000351.txt from 2001/01

From: George Kidder <gkidder@-----.org>
Subj: RE: [kl] Music Programs for PC??
Date: Thu, 11 Jan 2001 02:22:05 -0500

Oliver and all,

The bulk of the errors are in misidentifying the duration of the note.
Music fonts differ, of course, and with the rather old music I have used,
the most common error was the failure to get the note's duration correct.
This is especially marked with unbarred 8th notes, which are often written
as 16ths or quarters. Another common error was the failure to recognize a
half note as a note, leaving a blank space. Both of these sorts of errors
are detected by the automatic checking of the time of each measure, so they
show a blue flag on the measures which don't scan properly. Sometimes
other notes are missing, particularly outliers from a set of barred notes.
Almost never is there an error in pitch identification.

The process is not as seamless as Midiscan, in that you first produce a
TIFF or bitmap file and then use that as input.

As an example of the time taken, I just scanned in Handel's Trio Sonata in
Bb Major, for Oboe, Violin and Basso continuo realization. I used the
piano score version, with the other instruments in smaller type at the top.
Sharpeye did a reasonably good job on the different size staves, and I had
the whole thing into MIDI form in about 2 hours. I still have some work to
do on the MIDI, since for some reason the first page of * has the solo
instruments confused. Thereafter it went better, and there were often
considerable stretches that did not require any correction. I normally do
the final corrections "by ear", after playing it back and checking if
anything seems funny. Accidentals are sometimes misplaced by the OCR.

Guess you must be better with the numeric keypad than I am!

George

At 16:35 1/10/01 -0600, you wrote:
>George,
>
>I did a test several years ago with MIDISCAN to determine how much time it
>would actually save me in converting printed scores to Finale files. From
>the cleanest copy the time saving ended up being around 20% from that
>required to put in the notes from my numeric key pad. I decided that the
>reason there was so little to gain was that MIDISCAN made all kinds of
>mistakes requiring essentially a full manual check measure by measure of
>what had been created. The bottom line is that very little time was saved.
> There is very definitely a psychological factor here. The user of a more
>efficient program develops a confidence in that efficiency and then cuts
>back on the manual checks. What kinds of errors are made by Sharpeye?
>
>Oliver
>
>At 07:09 PM 1/10/01 -0500, George Kidder wrote:
>
>>I might also suggest you investigate "Sharpeye", which is a music OCR
>>program working with a flatbed scanner to produce a product which can be
>>converted into a MIDI file and then imported. It is not 100% accurate, but
>>it is better than the demo version of MIDIscan which I tried a few years
>>ago, and is much cheaper ($50). I am still trying to decide whether to
>>purchase it or not - probably will.
>>
>
>
>---------------------------------------------------------------------
>Unsubscribe from Klarinet, e-mail: klarinet-unsubscribe@-----.org
>Subscribe to the Digest: klarinet-digest-subscribe@-----.org
>Additional commands: klarinet-help@-----.org
>Other problems: klarinet-owner@-----.org
>
George Kidder
Bar Harbor, ME

---------------------------------------------------------------------
Unsubscribe from Klarinet, e-mail: klarinet-unsubscribe@-----.org
Subscribe to the Digest: klarinet-digest-subscribe@-----.org
Additional commands: klarinet-help@-----.org
Other problems: klarinet-owner@-----.org