Releases

4.3 Release

26-Nov-2013
Introduction
The release results from the adoption of Java Web Start to provide a single and consistent way to download / install / run Audiveris.
NOTA: This release should be the last one in the 4.x series. Development of Audiveris 5.x has begun, with the main objective to support even poor-quality scores.
Distribution Changes
  • Installation: Regardless of the underlying OS (Windows or Linux) and architecture (32-bit or 64-bit), Audiveris software is now made available via a single entry point on its web site. This takes care of all dependencies (Java runtime, C++ runtime, binary libraries, Tesseract languages, etc).

4.2 Release

7-Dec-2012
Introduction
The major features of this release were made possible thanks to the interfacing of new Google Tesseract C++ code (v3.x). The too few members of Tesseract project had no time to provide a clean DLL interface of their new version.
Kudos to Maxim Polijakovski for this long awaited bridge and for the configuration of Audiveris on provided OSes and machine architectures.
Many thanks also to Ryo Hang who took the initiative to address non-uniform gray images and came out with the prototype of an adaptive binarization filter.
Distribution Changes
  • Installation: Major OSes Windows and Linux and machine architectures x86 and x64 are supported via dedicated installers. The installers take care of all Audiveris dependencies (Java runtime, C++ runtime, musical font, etc).
    Nota: Support for MacOS has been dropped from the scope of this release to avoid further delays.
  • OCR: A few selected languages are pre-installed with Audiveris distribution (deu, eng, fra, ita). Additional languages can be supported by downloading the related trained data from the dedicated Tesseract web page.
  • NetBeans: A pre-populated nbproject folder provides NetBeans support out-of-the-box.
New Features
  • OCR: Tesseract V3.02 has been integrated in place of oldish V2.04 version. This much more powerful engine has led to a global redesign of text handling within Audiveris. There is now a dedicated TEXTS step which performs a layout analysis on each whole system image and transcribes the identified textual items. Note also that several languages can be selected at the same time.
  • Binarization: Extracting foreground pixels from the background has long been performed using a global threshold on pixel gray value. Even images with non-uniform illumination can now be processed with an adaptive filter which takes into account the neighborhood of the current pixel.
  • Glyph recognition: The major part of neural network input consists in moments which capture glyph key characteristics. Former Hu geometric moments have been replaced by ART moments (Angular Radial Transform, as used by MPEG-7) which are less sensible to noise.
  • Plugins: Audiveris MusicXML output can be "piped" to external softwares such as score editors or MIDI sequencers, through a flexible plugin mechanism. Consequently, these features have been removed from Audiveris application.
Bug Fixes
  • PDF input: Several free Java libraries have been tested (PDFRenderer for a long time, then JPedal and PDFBox) but none was really satisfactory. Hence support for PDF input is now delegated to a Ghostscript sub-process, a fully functional and perennial solution.
Other Changes
  • Doc: A comprehensive Handbook is now available from Audiveris web page, as well as the API JavaDoc of the current release. The former installation tab is now merged with the first chapter of the handbook.
  • Wiki: The online Audiveris Wiki contains detailed documentation about how to process each score of the set of examples available on MakeMusic/Recordare site. It is also used to gather information about evolutions being considered for Audiveris software.

4.1 Release

12-Oct-2011
Introduction
The main improvement in Audiveris processing stems from the use of filaments organized in a kind of grid. This grid serves as a referential for the following tasks, and provides an elegant approach for skewed or distorted images.
Distribution Changes
Several installation files have been published, all using the 4.1beta core name. This reflects the status of continuous development rather than stable release of the software.
New Features
  • Filaments: They are long glyphs representing the core of either horizontal or vertical lines (staff lines candidates and barlines candidates respectively). These filaments are formalized in natural splines, which are sequences of Bézier curves with continuity up to the second derivative.
  • Grid: The staff lines and barlines are connected into a grid of sometimes rather wavy lines. The grid itself is taken as the referential for all the other glyphs, whatever the potential skew or other distortion of the image, and thus saving the need for any pre-processing. Moreover, one can on demand easily build and save a "dewarped" version of the initial image.
  • Scale: Additional key informations are derived from run length histograms (jitter on staff line thickness and spacing, typical beam height, whether the image is music or not, ...).
  • Systems: The boundary between two consecutive systems is now a broken line, resulting from the incremental inclusion of glyphs into their nearest system.
  • Training: Besides full sheets taken as training samples, the user can select a mode that takes every manual assignment as a new training sample.
  • Symbols: The HEAD_AND_FLAG family of compound symbols no longer exists, thanks to an aggressive strategy in glyph split pattern.
Known Issues
  • OCR: We are still stuck to the old Tesseract version (2.04). The new Tesseract generation (3.x) has been out for more than one year now but still lacks a Java connection under Windows.
Other Changes
  • Time: All time values, such as offsets within a measure, are computed using rational values, which makes them independent of the score divisions value.

4.0 Release

24-Dec-2010
Introduction
The most prominent feature is the redesign of Audiveris display which combines sheet and score elements into a single view. This improves clarity and saves real estate.
New Features
  • Display: The main application window has been simplified. Only two views are now shown for each sheet: Picture (focused on input image) and Data (focused on items detected). We no longer have separate windows for sheet and score. The score elements are displayed in a translucid manner on top of the sheet glyphs they represent, in order to visually catch any discrepency. Separate voices can be displayed each in a specific color.
    Every other window (Log, Errors, Boards) can be displayed or hidden, and each individual board can be selected at will.
  • Font: Former symbol bitmaps have been dropped for the use of a TrueType music font (Stoccata.ttf then MusicalSymbols.ttf). This allows endless zooming of displays and printouts with no loss of quality.
    The font is even used to build artificial symbols used for initial training of the neural network.
  • Print: Ability to print the resulting score into a PDF file.
  • Multi-page: Multi-page images (using PDF or TIFF format) can be transcribed to multi-page scores in memory.
    A disk-based prototype version, using a map/reduce approach, allows to combine existing MusicXML pages into a single score.

3.4 Release

22-Aug-2010
Introduction
This release is focused on user interface, with drag n' drop features from a shape palette and a tight connection between the MIDI player and Audiveris display.
Distribution Changes
  • Libraries: All the external jars (23 as of this writing) needed to rebuild and/or run Audiveris are now provided in a dedicated /lib folder available in the download area. A developer can still pick up a newer jar version from the Internet.
  • Player: The XenoPlay MusicXML player has been replaced by a better player, named Zong!, still developed by Andreas Wenger.
New Features
  • Bench data: To allow the analysis of multiple batch runs, and compare the recognition efficiency, each sheet processing can log key data in a dedicated file. For the same purpose, time-out values can be defined for script or step processing.
  • Bar Lines: The user can now interactively assign / deassign a bar line that defines parts, thus recreating the systems from scratch.
  • Constants: All application constants can now be set from the CLI with the -option keyword. This complements the ability to set them from the Tools | Options UI menu.
  • Dots: Support for double dots, ability to assign the role of any dot (augmentation, repeat bar line, staccato, ...)
  • Horizontals: Horizontal entities (ledgers, endings) can now be forced or deassigned manually.
  • MIDI Player: The MIDI playback is now driven from a separate console window, borrowed from Zong! player.
  • OCR: Tesseract OCR is now available under both Windows and Linux.
  • Score: From a dedicated Shape palette, the user can Drag n' Drop a (virtual) glyph to either the score view or the sheet view, thus injecting entities directly into the score structure.
  • ScoreView: The zoom of the score view can now be adjusted at will, thanks to a slider and better symbol bitmap definitions. A next version will replace them with the use of Stocatta true-type font.
  • Time Signature: The user can now enter any custom time signature, defining numerator and denominator values explicitly.
  • Time Slots: Within a measure, the time slots are meant to gather notes that begin at the same moment in time. The user can now choose at the score level the policy for determining the time slots, either through stem alignment or through note head alignment.
  • Tuplets: 6-tuplets are now supported, as well as tuplets mixing beamed notes with other notes (flagged notes, rests, ...).
  • UI: A new board (Shape palette) is available. It allows drag n' drop for entity injection, easier navigation through shape ranges, and shape assignment by double-click.
  • UI: All boards now have an expand / collapse mechanism, thus allowing to save room in the column of boards.
Bug Fixes
  • Player: The Zong! Player is now more tolerant with respect to measure defects. It no longer throws an exception whenever the notes durations within a measure are not consistent with the measure expected duration.
  • Player Data: The data part of Zong player is now provided as a resource in a dedicated jar file, thus allowing the launching of Audiveris from any location of your computer.
  • Exception handler have been removed from all unitary tests, so that the results are clearly seen as successes or failures
Known Issues
  • Virtual Glyphs: For the time being, the (virtual) glyphs created by direct injection cannot be moved or resized once they have been dropped from the Shape palette to their target view. However, they can be deleted and re-injected (this workaround addresses a move but not a resize).
Other Changes
  • Images: Support for most pixel sizes.
  • Lyrics: Much better handling of lyric text pieces, with the ability for the user to enter extension sign or to split words with a space. The OCR can process several text lines as a whole, which often leads to better results.
  • Symbols: The symbols bitmap definitions (in the /symbol folder) have been refined with at least a 16-pixel interline definition, resulting in better display notably in score view.
  • Tiff: Images are forwarded to Tesseract OCR by memory, avoiding temporary files

3.3 Release

09-Jun-2010
Introduction
  • Release 3.3 is the first Audiveris release available on Kenai, a much more convenient environment than former java.net. So the source code is now handled by Mercurial (Hg) instead of CVS and JIRA is the issues manager.
  • The old Audiveris page on java.net now points to the Kenai site. Only the users mailing list has been kept on java.net.
  • With Sun / Oracle integration, java.net is announced to be re-implemented on top of Kenai infrastructure. So, these 2 Audiveris sites should merge one day.
Distribution Changes
  • Libraries: All the external jars (23 as of this writing) needed to rebuild and/or run Audiveris are now provided in a dedicated /lib folder available in the download area. A developer can still pick up a newer jar version from the Internet.
  • Player: The XenoPlay MusicXML player has been replaced by a better player, named Zong!, still developed by Andreas Wenger.
New Features
  • Bench data: To allow the analysis of multiple batch runs, and compare the recognition efficiency, each sheet processing can log key data in a dedicated file. For the same purpose, time-out values can be defined for script or step processing.
  • Bar Lines: The user can now interactively assign / deassign a bar line that defines parts, thus recreating the systems from scratch.
  • Constants: All application constants can now be set from the CLI with the -option keyword. This complements the ability to set them from the Tools | Options UI menu.
  • Dots: Support for double dots, ability to assign the role of any dot (augmentation, repeat bar line, staccato, ...)
  • Horizontals: Horizontal entities (ledgers, endings) can now be forced or deassigned manually.
  • MIDI Player: The MIDI playback is now driven from a separate console window, borrowed from Zong! player.
  • OCR: Tesseract OCR is now available under both Windows and Linux.
  • Score: From a dedicated Shape palette, the user can Drag n' Drop a (virtual) glyph to either the score view or the sheet view, thus injecting entities directly into the score structure.
  • ScoreView: The zoom of the score view can now be adjusted at will, thanks to a slider and better symbol bitmap definitions. A next version will replace them with the use of Stocatta true-type font.
  • Time Signature: The user can now enter any custom time signature, defining numerator and denominator values explicitly.
  • Time Slots: Within a measure, the time slots are meant to gather notes that begin at the same moment in time. The user can now choose at the score level the policy for determining the time slots, either through stem alignment or through note head alignment.
  • Tuplets: 6-tuplets are now supported, as well as tuplets mixing beamed notes with other notes (flagged notes, rests, ...).
  • UI: A new board (Shape palette) is available. It allows drag n' drop for entity injection, easier navigation through shape ranges, and shape assignment by double-click.
  • UI: All boards now have an expand / collapse mechanism, thus allowing to save room in the column of boards.
Bug Fixes
  • Many fixes, please refer to Hg log command
  • Player Data: The data part of Zong player is now provided as a resource in a dedicated jar file, thus allowing the launching of Audiveris from any location of your computer.
  • Exception handler have been removed from all unitary tests, so that the results are clearly seen as successes or failures
Known Issues
  • Player: The player raises an exception whenever the notes durations within a measure are not consistent with the measure expected duration, resulting in the unability to play any not-totally-correct score. This is to be fixed in the external Zong! player.
  • Virtual Glyphs: For the time being, the (virtual) glyphs created by direct injection cannot be moved or resized once they have been dropped from the Shape palette to their target view. However, they can be deleted and re-injected (this workaround addresses a move but not a resize).
Other Changes
  • Images: Support for most pixel sizes.
  • Lyrics: Much better handling of lyric text pieces, with the ability for the user to enter extension sign or to split words with a space. The OCR can process several text lines as a whole, which often leads to better results.
  • Symbols: The symbols bitmap definitions (in the /symbol folder) have been refined with at least a 16-pixel interline definition, resulting in better display notably in score view.
  • Tiff: Images are forwarded to Tesseract OCR by memory, avoiding temporary files

3.2 Release

19-Nov-2009
Introduction
  • Release 3.2 has been a long work, with a lot of new features. It exhibits a special focus on the integration of the Tesseract public OCR to handle all textual informations, and a significant redesign of key processing steps
Distribution Changes
  • Examples: Included 2 PDF examples (BrahWiMeSample.pdf and Dichterliebe01.pdf) by curtesy of Recordare. They demonstrate rather complex measures in terms of voices.
  • Application: We have switched to the Swing Application Framework, in order to use a standard approach for such tasks as: application lifecycle, resources, actions, session persistency,... all in an OS-independent way. This framework is bypassed when Audiveris runs in batch mode.
  • EventBus: This event bus from Michael Bushe has been added for notification of user selections, using specific user event types.
  • PDFRenderer: The PDFRenderer tool has been added to allow the reading of PDF images.
  • Tesseract: The opensource Tessaract OCR is the engine for text recognition.
  • Tesjeract: Access to Tesseract C++ software from Java via a JNI bridge.
New Features
  • Barlines: Systems building and Measure building have been redesigned, to allow easy assignment or deassignment of a measure barline by user interaction. The former BARS step is now split into SYSTEMS and MEASURES steps.
  • Drag'n Drop: Audiveris now supports drag'n drop features, dragging a image file to the Audiveris main window loads the sheet and perform a (customizable) default step.
  • Files: Files for Script, Score export, Midi can be specified via "Save / Save As".
  • Image: PDF format has been added to the collection of image formats that Audiveris can load, thanks to the PDFRenderer companion tool.
  • Midi: The user now has the ability to select and playback just a sequence of measures, besides the whole score.
  • Notations: All raticulations such as legato, staccato, accent, etc are recognized and translated.
  • Patterns: After automatic glyph recognition, a series of tests are gathered in various glyph patterns to handle specific cases (slurs, beams hooks, sharps, naturals, broken clefs, imposed time signatures, etc). Thus the PATTERNS step replaces the former CLEANUP step.
  • Parameters: All major score parameters are gather in a single dialog box for quick update.
  • Parts: Systems can now have different numbers of parts from one system to the other.
  • Score Layout: The user can dynamically change the layout of the score display, and switch between a vertical layout (like the original sheet) and a horizontal layout (more like a timeline).
  • Score Rebuild: The user has the ability to rebuild the score entities at any time. When dealing with very large systems, one can also disable the automatic score rebuild and perform the rebuild manually later when the modifications are completed.
  • Section Selection: Besides glyph selection, the user can now select one or several sections, thus building a glyph by hand. This allows the splitting of glyphs at section borders.
  • Copy/Paste metaphor: The user can copy a shape from an existing glyph and paste it to another glyph.
  • Similar Glyphs: The user can now browse through a sequence of glyphs "similar" to a selected glyph. This is helpful when dealing with similar cases all around a music sheet. The "similarity" is handled by a specific Linear Evaluator since a Neural Network is not linear at all.
  • Slots: Slots with insufficient margins are looked for, because they may indicate a too small slot margin value.
  • Steps: Some steps, the ones that follow the SCORE step, are flagged as not mandatory and can be re-performed at will. The user has bow the ability to reperform (almost) any step.
  • System Boundaries: The boundaries of any system area are now defined by broken lines, that the user can modify by mouse movements. This allows a more flexible definition of which glyphs belong to a system.
  • Text: A whole hierarchy of text items is now defined to precisely handle Lyrics as well as Titles, Directions and the like. The actual text content can be entered manually via the glyph board and automatically via the OCR mechanism.
Bug Fixes
  • Active Glyphs: Use of a reliable test for detecting which glyphs are active in a system.
  • Beams: They do not divide the duration of embraced notes when these ones are rests.
  • Message Log: The Log window now displays all messages, including the ones that were logged before the GUI was launched.
  • Local Browsing: The user is now able to navigate to the Audiveris documentation stored locally in the installation folder.
  • MusicXML Export: Fixed wrong order in elements for lyrics item, which triggered error messages in Finale. Also removed an xmlns:ns2 namespace which was legal but not accepted by Finale.
  • Vertex removal: The removal of a vertex sometimes led to a double removal attempt.
Known Issues
  • Injection: The user needs the ability to directly enter score entities of his own. To be addressed in the next release.
  • Parameters: Besides the generic constants mechanism (accessed via Tools | Options menu), we need a more user-oriented way to deal with global parameters. The current ScoreBoard is the first step in this direction.
Other Changes
  • Bus: Introduced the EventBus companion tool, to replace the proprietary Selection mechanism for handling the notifications of user selections.
  • MusicXML: Upgraded the export step to comply with the new JAXB binding of ProxyMusic.
  • Options: This huge tree / Table has been revamped for better usability.
  • Script: All user actions that may impact the resulting score are continuously recorded in a script in memory, and any such action is in fact implemented as an immediate execution of the ScriptTask instance.
  • Threading: Multi-threading is given much attention, using Brian Goetz JCIP annotations.
  • UML: Used the basic UMLet tool, which is very convenient for simple UML drawings found here and there in Audiveris doc.
  • Statistics: Added on Audiveris home page a link to the Ohloh statistics about this software. Cost is estimated a bit above 1 million $, just for the code ...

3.1 Release

28-Jan-2008
Introduction
  • Release 3.1 has focused on user interface and on Midi playback
Distribution Changes
  • Midi: The Midi features are provided via an external component, a Java application named XenoPlay from Andreas Wenger. The corresponding jar file is packaged with the other companions tools.
  • SwingWorker: To properly support the Mac platform, for which no Java 6 is available yet, we use the standard implementation of SwingWorker but provided as a stand-alone component.
New Features
  • Plugins: A knowledgeable user has the ability to register additional user actions, related to current Sheet or current Score. The initial version of score Player was implemented like this. Since then, all existing actions have been retrofitted to benefit from the decoupling brought by this plugin mechanism.
  • Player: A tight integration with Xenoage MIDI Player is provided. It enables to playback the whole score, or a range of measure, or just one selected measure using default or customized parameters. The user can enter MIDI parameters to the score, such as tempo, velocity and instruments. Interface with Xenoage player has been enhanced to allow callback from the player in order to move the score view in sync.
  • MIDI export: Besides the score player, we now provide the ability to export the current score to a MIDI file, still using Xenoage software.
Bug Fixes
  • Image Format: Audiveris now correctly handles black & white image formats for which a pixel is coded as one bit. Note however that such image must be converted on-the-fly from 1-bit to 1-byte format, which requires additional memory resources.
Known Issues
  • OCR: Audiveris is still lacking an OCR engine, but this should be fixed in 3.2 release.
Other Changes
  • MusicXML: ProxyMusic now handles only the latest version of MusicXML (2.0 as of this writing). The user can still use an external XSLT transform (using 'to11.xsl' sheet for example) if he wishes an older MusicXML version.

3.0 Release

29-Jul-2007
Introduction
  • This 3.0 release is a major release which makes Audiveris a complete music scanner.
Distribution Changes
  • XML binding: No more JiBX or Castor, we now use the standard JAXB 2 for all [un]marshalling tasks, including the score writing delegated to ProxyMusic.
  • Installation: To simplify installation, all companion jar files are stored in the /lib subfolder, with audiveris jar file refering them in its manifest. The only exception is JAI which needs to be specifically installed in JRE or JDK.
New Features
  • Translation: Implementation of translation on the base of extracted symbols.
  • Translation: Storing of the translation result in MusicXML format.
  • GUI: New window dedicated to score view, to more easily check the translation results.
  • Script: A scripting feature allows to record every action made on a given sheet, and to replay and augment the script at will.
Bug Fixes
A lot!
Known Issues
  • OCR: We are still looking for an OCR to handle lyrics and other written informations.
Other Changes
  • A lot!
  • Parallelism: we now take advantage of multi-processors when available, in the building of lag runs, and in the processing of system symbols.

2.4 Release

15-Jan-2006
Introduction
  • This is a release intended to cleanup major deficiencies before switching to the development of release 3.0 with brand new features.
Distribution Changes
  • XML Mapping: We've dropped Castor for JiBX. JiBX is reported more performant, especially at startup, since the binding is pre-compiled directly in the class byte code. Moreover, JiBX is the XML mapper used by MusicXMLrep which we'll use to handle the mapping to MusicXML format. So instead of the Castor environment, the JiBX one must now be installed (see the Installation Manual for details).
  • Tests: The distribution archive now comprises the source code for unit tests, based on the JUnit framework.
Bug Fixes
  • Memory: At last, the origin of memory leak was discovered: it originated in Swing component handling an HTML text. So the workaround has been to simply drop HTML used by FilterBoard in favor of standard JGoodies form fields of CheckBoard.
Known Issues
  • Java: There are 4 compiler warnings related to generics.
Other Changes
  • Math: The Line class is now better defined as an interface with a BasicLine implementation which can be used whatever the orientation of the Line.
  • Java: Consistent application of an inner Swing component rather than direct derivation from Swing, except in obvious specializations. Main improvement is visible in Javadocs that used to be filled with Swing inherited members.

2.3 Release

10-Nov-2005
Introduction
  • This release is rather consistent in terms of extraction ratio, thanks to the adoption of CLUTTER and STRUCTURE shapes, and the focus on a single evaluator. Unless a major bug needs to be fixed, this release should conclude the 2.x series focused on symbol extraction.
Distribution Changes
  • Properties: the user property files, used respectively for build and run phases, are now located within a ".audiveris" folder in ${user.home} directory.
  • Icons: The "Musette" font is no longer used, since it lacked too many symbols. Instead, we now use our own drawing of icons for all shape symbols, using a simple text format for storing icon definitions.
  • Training: A set of representative glyphs is provided as a "core", should the user need to retrain the glyph evaluator. This core is a carefully selected subset of the whole base of glyphs.
New Features
  • Extraction: Added two specific Shapes, known as CLUTTER and STRUCTURE, so that evaluator can differentiate between a valid symbol and symbols that either should be discarded (Clutter) or kept for later leaves retrievals (Structure).
  • Extraction: Got rid of the Regression evaluator, and organized all recognition steps on the sole use of Neural Network evaluator.
  • GUI: Simplified and rationalized the mouse events related to selection, lasso, etc... It is now possible to incrementally select several glyphs and process them globally. This eases the user actions when manual assignment is needed.
Bug Fixes
  • GUI: Removed duplications of mouse events.
Known Issues
  • Memory: The memory leak (see previous version) is still to be fixed.
  • Java: There are still 5 compiler warnings related to generics, no harm, but they should be sorted out.
Other Changes
  • Icons: To define custom icons, there is a new stand-alone tool (SymbolRipper) which allows to store a symbol representation drawn with help of any suitable installed font.

2.2 Release

31-Aug-2005
Introduction
  • This is a release still focused on improvement of existing features rather than functional extension.
  • In particular, the attempt to handle clutter as a recognized glyph shape in the neural network (to avoid false positives) has been postponed to next release.
Distribution Changes
  • Tools: All the companion tools have been upgraded to their latest version, and the Installation manual updated accordingly.
  • Documentation: The whole set of documentations, including the JavaDoc API, has been organized under the same umbrella to allow convenient access to any piece from the documentation banner.
New Features
  • Image: Added the conversion from color image to gray-scale image. This operation is launched automatically when a color image is detected at load time.
  • Glyphs: Added a Deassign button in the Glyph board, with action depending upon the glyph at hand (for example deassigning a stem has important consequences, more than for a standalone symbol).
Bug Fixes
  • GUI: No more toggle button is active for the Bars step, since there were no specific Sections to handle.
  • Documentation: Fixed the use of CVS tags in java files (for version information) which were mistaken with real CVS tags in related javadoc HTML files
Known Issues
  • Memory: There is still a memory leak, since even after current Sheet has been closed, there is still a Sheet instance referenced with Score, related views, etc... Workaround if several Sheets have to be processed interactively is to exit the application after each processing.
  • Documentation: Because of JavaScript building of doc frames, the "Back" and "Forward" commands trigger duplicate frame loads. Function is OK, but this flickering should be avoided.
  • Java: There is still half a dozen of compiler warnings related to unchecked generics assignments, to be cleaned in a next release.
Other Changes
  • Verticals: Managing of stems is done as much as possible like other glyphs, since managing separate lists in synch was too much error-prone.

2.1 Release

03-Jun-2005
Introduction
  • This release is a small improvement on the initial 2.0 release, with major focus on user documentation, especially the Operation manual.
New Features
  • Added documentation environment (installation, operation, design, implementation).
  • Added a nearly complete Operation manual.
  • Added the handling of a Core base, besides the Whole glyph base, to speed-up the training of the neural network.
  • Added the use of manifest parameters to automate the handling of tool name, version and build reference.
  • Added in the Regression evaluator the ability to measure distance to a given shape, as a way to select members for the Core base.
Bug Fixes
  • Fixed a bug in resource names due to improper use of backslashes.
  • Fixed the loading of unsupported image formats, by simply detecting the event and notifying it to the user.
Known Issues
  • Audiveris should be able to convert an unsupported image format to the desired one (grey scale of 256 values).

2.0 Release

05-May-2005
Introduction
  • This release is the result of the strong need to formally publish the Audiveris software, even though there are still many areas with lots of room for improvement. Publishing the software, with source code and a minimal documentation, was necessary to let other potential developers contribute to this project.

1.0 Release

ca. 2000
Introduction
  • Prior to Audiveris, the author had built an OMR prototype called OptiMusic and developed using the Ada technology.
  • OptiMusic could be considered as version 1.0 of Audiveris.