Hi,
did anyone know if scanning of sheet music and translating in Notes is planned for Dorico in future?
I can’t imagine it’s a remotely high priority for Dorico’s developers, given there’s existing software specifically for this purpose that exports fairly good XML, and Dorico’s import is good.
Of course, by that token there’s existing notation software and existing DAW software so you may want to entirely disregard this post!
The OCR’ing of music is akin to teaching a pig to sing, the results aren’t great and it annoys the pig. It took years for Neuratron to develop PhotoScore. Now, I use PhotoScore, I even love PhotoScore; it saves me time. But there is no denying it is far from perfect despite all the years of top-notch development. I can’t speak for them, but for the Dorico team to even remotely think about getting involved in this hornet’s nest would seem foolish to me when other software does does this and their development time on creating a new scoring program is limited. So I happily use PhotoScore.
Ok, I have to ask - how do we know that the pig is annoyed?
Sorry, couldn’t resist…
… a pig farmer knows …
I’m a Photoscore user myself, and I second everything Claude has written in his first post.
I wish Neuraton had a forum like this one…
It is an area that seems ripe for innovation though. The development of machine learning libraries like Tensor Flow, combined with the availability of a huge corpus like IMSLP might mean that a radically different approach is possible? Who knows.
I will ask no more questions on this topic…
It’s at times like this I miss the like button … 
Has anyone testet SharpEye2 ?
I have downloaded a Testversion of the programm.
The first results with easy notation looks ok, but i have to test it further.
And it only takes files in BMP or TIFF. if you have already pdf you must convert them.
And the actual version 2 is from 2006!
That looks like it is not longer supported?
I believe the engine from SharpEye is now incorporated into PhotoScore, and its author, Graham Jones, is no longer working on it.
I hope this is not seen as spam, as I am suggesting one of our programs as solution. 
We have just released ScanScore as a standalone license.
ScanScore should work fine with dorico via musicxml. Maybe that could be interesting for you: https://scan-score.com/en/
It was integrated into our latest version of FORTE and a lot of our users are happy with it. However Music OCR is not easy and there are cases when results are not satisfactory. But I think this is the case with every one of those programs.
In a few weeks we will release the pro version with more features.
Disclaimer: I am the CEO of the company that develops and sells ScanScore and FORTE. I think this is relevant to the topic and might help. Please remove this post, if you see it as inappropriate.
Dear SebastianLugert,
You should make it clear in your post that your software is windows only, and by what I see, it is still under development (which is not a bad thing, it means it’s alive, but still… worth noting it).
I’ll try it when it works on my machine.
Other options:
Musitek’s Music-to-XML app is cheap (for music OCR) at $99, but will usually produce a spikey dog unless you buy the full SmartScore app that lets you edit and correct the results before export. People report that using the full app does produce good results, but the cheaper option is fairly unusable.
Myriad’s PDFtoMusic Pro works well at producing XML from vector PDFs (i.e. exported from other notation apps), but cannot read scanned images. This is because it looks at the vector data for lines, font characters, etc, to produce the score. (Only the Pro version exports XML, which is a bit inexplicable.)
You should make it clear in your post that your software is windows only, and by what I see, it is still under development (which is not a bad thing, it means it’s alive, but still… worth noting it).
Sorry for not mentioning. Yes this is important.
We are still working on our Pro Version 1.0 which will be released in a few weeks.
We are planning to release a Mac Version maybe end of 2019.
I used that software many years ago. I just visited their website. It seems they are making some nice advances lately.
Have you tested their new Audioscore product?
And just for the record, people should take a look at PDFtoMusic from Myriad. This product reads PDFs that were written directly by a music notation product (not scanned) and usually does an impressive job of putting it all back together in a usable XML file. There are some complex cases where things get pretty mixed up between PDFtoMusic and Dorico, but most of the time, it is a big time saver if you have a directly written PDF to start with. I have used that to convert brass quintet arrangements to trombone quartet, for example. Sometimes I can do a project like that in an hour or two.
http://www.myriad-online.com/en/products/pdftomusic.htm
PDFtoMusic is a great tool for straightforward scores but it can make a complete hash of anything complex.
This is “sort-of-recognisable as the same score,” but useless for anything else:
Yes, I have seen some similar things. And in my case, the Myriad output displayed much, much more like the original when imported into Finale. I think it is understood that Dorico’s XML import is a work in progress. And it seems the Myriad software does some unnecessarily goofy things when writing the XML for a complex piece.
Fortunately I never deal with any notation as complicated as your example. I always wonder if humans even try to play that type of stuff literally.
The problems aren’t Dorico’s XML input. PDFtoMusic pro just can’t find the notes, or judging by what you see if you go into edit mode inside it, it can’t find beams that overlap staff lines.
Yes, humans have played that (or more pedantically “a human”, since it’s a solo piece) . It’s been performed live several times, (the total duration is about 1 3/4 hours,) broadcast complete more than once, and recorded. The recording is pretty close to “literal.”
It’s not “modern music” either - it’s almost 100 years since it was composed.
Rob, judging only from the snippet you posted, if this organ music were performed in any remotely reverberant space it would run together and be a mush of sound would it not? If not why not?