optical recognition for input

[James B. Wilkinson]

I'm looking at getting an OCR program to recognize scans in pdf files and 
create MusicXML files representing the contents. I would then use the 
musicxml2ly or perhaps some other script to translate the MusicXML to Lilypond 
files, I imagine that some of you are already doing this, and I wonder whether 
you could give me or point me to some tips as to what software to use and how 
to go about it.

Using MIDI files as the intermediate representation may also be a possibility. 
I'd appreciate hearing about that, too.


thanks