Download

Get the latest version of OSRA:

OSRA 2.1.3 source code (free)
OSRA 2.1.3 for Linux (purchase)
OSRA 2.1.3 for Windows (purchase)
OSRA 2.1.3 for OSX (purchase)

Download patched GOCR here: patched GOCR source code
Download patched OpenBabel here: patched OpenBabel source code. Use openbabel-3-0-0-patched.tgz with OSRA 2.1.1 or later.

Changelog

• OSRA 2.1.3 - Fixed aromatic bond detection and coordinate box calculation.
• OSRA 2.1.1 - Improved PDF processing. Updated poppler to 0.73.0. Updated openbabel to 3.0.0. Improved reaction recognition.
• OSRA 2.1.0 - Improved recognition for PDF files, recognition of polymer structures.
• OSRA 2.0.1 - Improved precision (filtering out false positives).
• OSRA 2.0.0 - Significantly improved recognition rates. Added recognition of Iodine, wavy bonds. Completely modified confidence function (values not compatible with the earlier versions).
• OSRA 1.4.0 – New features include reaction recognition, better bounding box coordinate detection, and an update to Accelrys Draw plugin to handle reactions.
• OSRA 1.3.9 – Bug fix and code cleanup release. Added single molecule MOL format output, improved Java (JNI) interface, significantly improved Accelrys Draw plug-in (James Jack). Added osra-pdf script for more thorough PDF processing.
• OSRA 1.3.8 – New capabilities include adaptive thresholding for low-light, low-contrast images, output of structure box coordinates, output of InChI, InChI-key or SMILES as SD file properties, new version of Symyx Draw plugin (now requires Symyx Draw 4.0), and the library version of OSRA which is possible to operate from Java through JNI. All OCR engines - OCRAD, GOCR, and Tesseract - have been updated to the latest available version. Note that newer versions of Potrace and TCLAP are also required.
• OSRA 1.3.7 – Thanks to the contributions by Dmitry Katsubo OSRA now can use the standard "./configure; make; make install" build process. Recognized Markush labels (R1, R2, R3...) are now added as atomic aliases in SD files. Added optional support for Tesseract and Cuneiform OCR libraries. Added additional processing for low scan quality documents (-j command line option).
• OSRA 1.3.6 – Updated superatom dictionary, added an option to print out average bond length as measured from the image, modified table detection routine, added an adaptive filter for images with too long or too short bonds.
• OSRA 1.3.5 – Improved memory handling and file read/write error handling. Added command-line option to save output to a file. Added call to InitializeMagick() (required for GraphicsMagick v1.3.8). Added multi-threading support (OpenMP) for processing many-page documents (Linux/Unix only, does not yet function on Windows). Added command-line option for saving output to a file.
• OSRA 1.3.4 – Improved table recognition. Fixed a bug affecting JPEG image processing on Windows. Fixed a bug affecting double bond detection on Windows. Added optional pre-processing with Unpaper routine (-u command line option). Added support for OCRAD v0.19.
• OSRA 1.3.3 – The tables (boxes) around the structures are detected and removed prior to processing, improving accuracy and speed. Added -R (--rotate) command line switch to rotate the image. Modified debug output (-d option) to show the output from superatom dictionary. Made debug output more consistent.
• OSRA 1.3.2 – Significant speed up achieved (by a factor of 2-3x) by replacing ImageMagick libraries with GraphicsMagick. Fixed some issues with slowdowns occurring when using Windows executable.
• OSRA 1.3.1 – Improved speed and various bug fixes. PDF processing now honors and -r (--resolution) command line options. You can get higher quality results (at the expense of slower speed) running with the following command line: osra -r 300 -f sdf file.pdf. Also, you can see the page number for structures from PDF documents with -e (--page) option.
• OSRA 1.3.0 – OS X install package - one click to install OSRA on a Mac. A [Plugin_for_ChemBioDraw]. Better recognition of high resolution images (above 300 dpi). Improved Symyx Draw plugin.
• OSRA 1.2.2 – Superatom labels can now be edited by users – superatom.txt contains the SMILES strings for each recongized label and spelling.txt contains spelling variants of every label for cases where OCR engine is not reliable. Please note that the dependencies have changed – OCRAD v0.18 is now required and RDKit support is temporarily suspended.
• OSRA 1.2.1 – There is a windows installer which automatically installs a plugin if Symyx Draw is present. It also detects and auto installs if necessary Ghostcript and GflAx libraries. Improved speed (up to 30% increase) and double and triple bond detection.
• OSRA 1.2.0 – Page layout analysis algorithm completely re-written. Added plugins for integration with several popular molecular editors: BKChem, MolSketch, Symyx Draw, and PipelinePilot are now included with Windows zip archive.
• OSRA 1.1.0 – Added SD file format output, improved wedge bond detection.
• OSRA 1.0.1 – Minor bug fixes. OpenBabel v2.2.0 or svn snapshot of RDKit are recommended with this version.
• OSRA 1.0.0 – Significant update of the recognition engine. Simplified built instructions. Please note that the dependencies have changed since the previous version.
• OSRA 0.9.9 – Build system upgraded to allow linking to the newer versions of GOCR (v0.45).
• OSRA 0.9.8 – Added recognition of old-style aromatic rings with heteroatoms.
• OSRA 0.9.7 – Improved recognition of color and low-res images.
• OSRA 0.9.6 – Introduced automatic resolution detection.
• OSRA 0.9.5 – Source code modified to facilitate compiling with MinGW for Windows platform.
• OSRA 0.9.4 – added old-style benzene ring recognition
• OSRA 0.9.3 – added rudimentary formal charge recognition
• OSRA 0.9.2 – improved handling of hash and wedge bonds
• OSRA 0.9.1 – slightly improved handling of 72dpi color images
• OSRA 0.9 – original public release