Optical Music Acknowledgment.

Uploaded on:
pictures of music (8bit dark) verses (first lines of verse and ensemble) ... Suitable OCR for verses not found. Business OCR frameworks are frequently insufficient for non ...
Slide 1

Optical Music Recognition Ichiro Fujinaga McGill University 2007

Slide 2

Content Optical Music Recognition Levy Project Levy Sheet Music Collection Digital Workflow Management Gamera

Slide 3

Optical Music Recognition (OMR) Trainable open-source OMR framework being developed subsequent to 1984 Staff acknowledgment and expulsion Run-length coding Projections Lyric evacuation/classifier Stems and notehead evacuation Music image classifier Score recreation Demo

Slide 4

OMR: Classifier Connected-segment examination Feature extraction, e.g: Width, tallness, viewpoint proportion Number of openings Central minutes k-closest neighbor classifier Genetic calculation

Slide 5

Overall Architecture for OMR Image File Staff expulsion Segmentation Recognition K-NN Classifier Output Symbol Name Optimization Genetic Algorithm K-nn Classifier Knowledge Base Feature Vectors Best Weight Vector Off-line

Slide 6

Lester S. Demand Collection

Slide 7

Lester S. Demand Collection North American sheet music (1780–1960) Digitized 29,000 pieces including "The Star-Spangle Banner" and "Yankee Doodle" Database of: content list records pictures of music (8bit dim) verses (first lines of verse and theme) shading pictures of spread sheets (32bit) http://levysheetmusic.mse.jhu.edu

Slide 8

Digital Workflow Management Reduce the manual intercession for huge scale digitization ventures Creation of information archive (content, picture, sound) Optical Music Recognition (OMR) Gamera XML-based metadata writer, lyricist, arranger, entertainer, craftsman, etcher, lithographer, dedicatee, and distributer cross-references for different types of names, nom de plumes forms of names and subject terms Music and verse web crawlers Analysis toolbox

Slide 9

The issue Suitable OCR for verses not discovered Commercial OCR frameworks are regularly deficient for non-standard reports The business sector for particular acknowledgment of authentic records is little Researchers performing report acknowledgment frequently "re-imagine" the fundamental picture handling wheel

Slide 10

The arrangement Provide simple to utilize devices to permit space specialists (individuals with particular learning of a gathering) to make custom acknowledgment applications Generalize OMR for organized reports

Slide 11

Introducing Gamera Framework for production of organized archive acknowledgment framework Designed for space specialists Image preparing devices (channels, binarizations, and so on.) Document division and examination Symbol division and grouping Feature extraction and choice Classifier choice and combiners Syntactical and semantic investigation G eneralized A lgorithms and M ethods for E nhancement and R estoration of A rchives

Slide 12

Features of Gamera Portability (Unix, Windows, Mac) Extensibility (Python and C++ modules) Easy-to-use (specialists and software engineers) Open source Graphic User Interface Interactive/Batchable (scripts)

Slide 13

Scripting Environment (Python) Automatic Plugin Wrapper (Boost) Architecture of Gamera Graphic User Interface (wxWindows) Plugins (Python) Plugins (C++) GAMERA Core (C++)

Slide 14

Example of C++ Plugin/Number of pixels in lattice #include "gamera.hh" #ifdef __area_wrap__ #define NARGS 1 #define ARG1_ONEBIT #endif utilizing namespace Gamera; layout <class T> feature_t area(T &m) { return feature_t(m.nrows() * m.ncols()); }

Slide 15

Example of Python Plugin/This channels a rundown of CC articles import gamera def filter_wide(ccs, max_width): tmp = [] for x in ccs: if x.ncols() > max_width: x.fill_matrix(0) else: tmp.append(x) return tmp

Slide 16

Gamera : Interface (screenshot in Linux)

Slide 17

Gamera : Interface (screenshot in Linux)

Slide 18

Histogram (screenshot in Linux)

Slide 19

Thresholding (screenshot in Linux)

Slide 20

Thresholding (screenshot in Linux)

Slide 21

Staff evacuation: Lute sheet music

Slide 23

Classifier: Lute (screenshot in Linux)

Slide 24

Staff expulsion: Neums

Slide 25

Classifier: Neums (screenshot in Linux)

Slide 26

Greek case

Slide 27

Conclusions Gamera permits quick improvement of space particular archive acknowledgment applications Domain specialists can alter and control all parts of the acknowledgment procedure Includes a simple to-use intuitive environment for experimentation Available on Linux, OS X, and Windows

Slide 28

Projections X-projections Y-projections back

View more...