Interest survey for including a scanning/OCR feature in Notation Composer -reply here

[rrayner]

Okay, Reinhold, interesting expansion concept. I like it. Why raise the price Composer for folks who wouldn't be using the scan feature. Brilliant!

So, now that you have shed a little more light on your approach, let me revise my views. The current Amazon price for Neuratron PhotoScore Ultimate 8 is $229.99, which is misleading, because from what I know, you also have to have Sibelius in order to get any kind of flexible editing of the scanned music.

So, if you add the $504 cost of Sibelius, you would have to pay $733.99 to get an almost workable capture -- for folks who would like the scan feature in Composer, that gives you a very wide margin between your Composer cost of $98.99 and Photoscore/Sibelius cost of $733.99.

As far as I am concerned, the $229.99 I paid for Photoscore is NOT worth it. I use it very little. I have some simple lead sheets that I will try again on, just to see if more frequent use makes me more proficient with it.

I hope this helps.

Ralph Rayner

[Reinhold H.]

Ralph, thanks for your comments.

The scanning quality of the concept would be exactly the same as the scanning quality of SharpEye 2 which uses the Liszt OCR scanning engine.
The technical concept basically is to export a MusicXML file by the Liszt OCR engine and import this MusicXML file into Composer.

There is a long thread about the best scanning OCR engine in the forum of the MuseScore folks. Unfortunately (or fortunately ) it is in German language where people contributed who did intensive testing and validations. Please find the thread here. SharpEye sounds OK and the testing that I did was OK, too for the music sheet that I used. I can recommend to try the SharpEye 2 trial version which you can download from the Visiv page. If somebody tries it, just compare the scanned score to your original music sheet and let us know what you think about the scanned quality.

From a technical platform perspective the beauty using the Liszt OCR engine is that we can perfectly use it across all platfroms which we support from Windows XP to Windows 10, Mac and Linux.

Reinhold

[rrayner]

Hi Reinhold,

I hope this reply won't clutter this thread with unwanted data, but I just tried another Photoscore 8 to Sibelius 7 to Composer capture, and I thought you might be interested in the results.

Included are the .pdf file (Internet), the .mid file (exported from Sibelius), and the .not file (Composer). I could not upload the .opt file (from Photoscore), but I don't think that is too important.

As you will see, the bar is not set very high. As you are considering the SharpEye approach, adding this update to this thread may be meaningless -- just giving you some insight into your competition.

Ralph Rayner

[Reinhold H.]

Ralph, Thanks for sharing.

Those files are important in order to validate the scanning results. PhotoScore also uses the SharpEye's engine with some enhancements. I took your PDF file, convert it to bmp and ran it through SharpEye. SharpEye is not able to properly scan the treble piano part (2nd part). The first upper part is accurately detected. This finding is important.
The scenario having the MIDI file in the chain makes it worth due to the midi file nature. So, the result here is - the SharpEye engine cannot scan this PDF file properly.

Reinhold

[rrayner]

Reinhold - are you interested in seeing any more Photoscore Ultimate to Sibelius to midi files? I am currently working on a couple of Recorder Duets and using this procedure to save time in Composer.

I could include them in this thread or send them in a separate email if you think they would be of any help. I am scanning pages from a fake book. I am going through the work on my own anyway, so there would be little effort to send them to you. Let me know.

Ralph Rayner

[dj]

Hi, guys and gals:

Replying to your request about scanning quality, Reinhold, it's been a long time since I did any serious scanning and the copy of Sharp Eye that I have is very old.

What I do remember is a lot of editing of each scan's resulting page (.MRO files) to get things more in shape, and then assembling the resulting files into one .mid file.

I've gone back and found the archive of the show that I used Sharp Eye to assemble and I've attached two files here. The .mid file is the result of the scanning process, after considerable hand editing, and the .not file is the final product. This, by the way, was scanned from a printed score.

And, the cost is a lot. I think maybe the high cost of music scanning software reflects the relative demand for it: when you need the capability, you really need it, but have to be prepared to pay for "niche" software.

Hope that helps.

David

[kknollman]

Hello.
It looks like this thread has not had a reply for several months, but I want to add a few notes from my experience. I had purchased PhotoScore quite some time ago to pair up with my original version of Composer, but my experience with it was very unsatisfactory. I remember it taking more time to edit the conversion than to create the score in Composer manually. I also tried SharpEye several years ago, but found it less satisfactory than I had expected.

Two years ago I purchased Capella-Scan, and find this to have a very satisfactory success rate recognizing scores. Where it senses errors by improperly recognizing some notation, it drops various color coded hints into the on-screen notation to aid in manually correcting recognition errors. The cost is in the $250 range.

I input pdf scores into Capella, and although I currently output them as midi files because that is the only format I can input to Composer, it does output to MusicXML, and I am anxiously awaiting the capability to input that format into Composer.

-Ken