Closed noahbaxter closed 5 years ago
most of this information is superfluous, the exception being the line attribute. As a result, JSOMR now contains staff information as well glyph information. Fixed in https://github.com/DDMAL/heuristic-pitch-finding/commit/c5c1a16097d5828a9cf5ede09e09c0e1664f05e1
What do you mean superfluous? By the time the MEI file leaves the pitch finder, it should contain all the information about the score (except for the text info).
Let me try to explain this better. The mei files that are currently generated contain, besides the simple document structure, staves and glyphs with bounding boxes. As this is enough information to be rendered accurately by Neon, any additional page layout information is unnecessary for our current purposes.
Should there be any more layout information than bounding boxes?
These mei files only ever contain a single page, so page breaks pg
is unnecessary.
Multiple staves are being rendered per page as opposed to one continuous one, so staff breaks sb
are unnecessary.
Divisions are just another type of glyph, and so if/when the classifier starts recognizing different division types, it'll already be as a glyph within a staff.
Finally, nonpitch attributes are being generated by neume type with the nc structure. Based on the neume name alone, it can be determined whether a nc is a punctum, inclinatum, or ligature and so these tags are added during mei generation.
The only information that seems necessary right now are staves, neumes/ncs + pitches, eventually letter/word glyphs, and bounding boxes for each.
The new CNN-based pitch finding and the MEI mapping tool will provide all the information necessary to handle these kinds of edge cases and rare types of neumes.
There is missing information about the following elements:
line
attribute instaff
orclef
element, representing number of staff linespb
, orsb
divisions
liques
,tilt
andcon