SharpEye Version History
Optical Music Recognition (OMR) software for RISC OS
Known problems in program
(Updated 18.3.00) There is a problem transferring files directly to Rhapsody 4. Sometimes R4 exits, leaving the mouse constrained to centre-screen. (Pressing <F12>, <Return> clears this.) Sometimes the hourglass stays on, and you have to press <Alt-Break> to quit R4. It seems to happen only when there is a blank score open in R4, and you drag directly from SE to this window or the icon bar. This appears to be a bug in R4, hopefully fixed in version 4.09 of Rhapsody 4.
The work-around for this problem is the obvious one: save R4 files from SharpEye to disc and then load them into R4, instead of dragging directly.
Changes to SharpEye Music Reader: most recent first
Changes 1.27 (20 March 2003)
First fully 32-bit-compatible version; now runs on RISC OS 5 (Castle IYONIX pc) as well as 26-bit RISC OS 3 & 4 systems. The latest System modules are now required for use with 26-bit machines (obtain the modules from Castle's site).
- New RISC OS 5-style application sprites for appropriate systems;
- New high-resolution pointers for all systems;
- Small interface improvements.
Changes 1.21 to 1.26 (17 September 2001)
Fixed bug in MIDI export (error message "...makemidi.c592").
Bug fixed. Rare crash in OMR engine during 'Analysis' stage, happened when very skewed staves were chopped off at image edge.
Stave location failed on some images if the staves were very close together. Now fixed.
In cases where simultaneous up and down stems are used in music for one instrument to 'stretch' a short beamed note, SharpEye did not output the correct duration for the note in MIDI. Now it should.
Changes 1.19 to 1.21 (3 December 2000)
In rare circumstances (when staves were at slightly different angles), lyric lines were sometimes duplicated. Should now be fixed or at least even rarer.
Fixed bug in the stave location code which meant the change I made in 1.19 sometimes made things worse, not better.
The 'User scan' menu in TWAIN's interface did not work properly (the submenus did not open). This was due to an interaction between the Toolbox and TWAIN which I do not fully understand, but I hope I have fixed it.
Changes 1.19 to 1.17 (5 October 2000)
In unusual circumstances a stave with beams obscuring staff lines could be missed. Now improved. [Technical: Added earlier weed out of candidate stave locations.]
Recognition now distinguishes between 'normal' double bar lines (two thin lines) and final double bar lines (thin line, thick line).
Improved stave location in cases where there are two sizes of staves which are very close. Occasionally complete staves were missed in such cases. This should be rarer now.
A couple of bugs relating to whole rests were fixed. A whole rest was taken as filling a bar in 2/1, 4/2, 8/4 time, and was not taken as filling a bar if it occured in the first bar of a score in 3/2, 6/4 time.
Bug fixed. Recognition engine failed on a score with rests but no notes.
Changes 1.15 to 1.17 (3 June 2000)
Made the <Backspace> key work as the <Delete> key.
Bug fixed. If there were several clef changes in a stave, and the 'remove repeated clefs' option was set for R4 export, then some clefs which should not have been removed were removed when making R4 format.
Bug fixed. If two staves were joined at the start of the score, and the green arrows were used to make the upper of these become the bottom stave, an inconsistent file could be produced, which caused SharpEye to crash when loading.
Bug fixed. In rare circumstances, the recognition engine crashed during reading text. [Technical: in readline_readline().]
Bug fixed. If you tried to delete a rest while dragging it with the mouse, SharpEye crashed.
Bug fixed. If you clicked on the tie tool at the extreme left, you got the vertical alignment tool instead.
[Technical] Bug fixed. *.mro reading routines did not correctly ignore comments. Only affects debugging versions of *.mro files.
Changes 1.13 to 1.15 (1.14 was Windows only)
Bug fixed. In rare circumstances, the recognition engine crashed with a divide by zero error during reading text. (Fix in findwordrecogcost().)
Bug fixed. It was possible to select a rest and notes in the music editor, then choosing a tool (e.g. tenuto) that is inappropriate for rests caused a crash.
Can now use the green arrows between systems to reorder staves, useful for dealing with irregular systems.
The stave spacing (of the first system) is now output to R4. Also make more room at top and bottom of the system.
Various technical changes to make porting the recognition engine to Linux easier.
Changes 1.12 to 1.13
Implemented more sophisticated test for when staves are joined to make systems; there were occassional failures, should be fewer now.
Bug fixed. Sometimes notes tied across bar lines caused incorrect MIDI to be generated. (I thought I'd fixed this in 1.11 but in certain situations it was still wrong.)
Bug fixed. Sometimes it was not possible to add ties to a chord to tie it to a chord in the next system.
Changes 1.11 to 1.12
Bug fixed. If there is no time signature at the start of a score, SharpEye defaults to 4/4. In this case, the MIDI tempo was ignored. Now insert an explicit 4/4 time signature with tempo into the MIDI file.
Pause signs over rests are better positioned when rests are not within stave.
The text OCR used code 132 for reject characters; now use
Bug fixed. The position of pause signs in Rhapsody 4 files was wrong.
Changes 0.19 to 1.11
(Note: 0.19 was the last beta version before release. I am jumping to 1.11 in order to keep the numbering in line with the Windows version which was released in May. I think this a good idea...)
If the scan was saved, but then deleted or moved, you could get two error messages. Now fixed.
Made closing the music window work according to the RISC OS style guide, i.e. get a Discard, Cancel, Save box, and made Adjust-close open the Filer window for the file. Closing the window and choosing Discard loses the data.
Scrap directory now deleted when SharpEye exits.