Fixed segfaults in groupRedraw and chain superatoms
Updated metrics spreadsheet
Added chain.txt to osra.nsi installer
Added metrics spreadsheet
Added Oxygen to the list of organic non-carbon atoms for mol stats
Relaxed limit on average bond length vs real font height
Do not calculate number of small angles if chains are present
Added superatoms from USAN
Added superatoms from uob
Finished uspto super atoms and chains
Working on adding chain superatoms from uspto
Working on adding chain superatoms from uspto
Added superatoms from JPO
Added chain.txt map
Started implementation of 2-connected superatoms - CH2, CH2CH2, etc.
Version bump to 2.2.3
Fix for extend terminal bond to label
Version bump to 2.2.2
Fixed a potential buffer overflow for mcdlutil
Improved remove zero bonds
Improved collapse double bonds function
Recognition improvements
Improvements for Markush structures
The changes in GOCR are concerned not with threading support but with building it as a library. You are welcome to try building from an unpatched source of course.
Thanks. Is it still possible to use OSRA with unpatched GOCR? Are the changes in the patched version substantial, or just concerning threading support?
The recommended way to build OSRA from source is to use patched GOCR source code from https://sourceforge.net/projects/osra/files/gocr-patched/ When compiling GOCR please make sure to run "make libs" to actually build GOCR library. By default only gocr executable is built.
Build failure with GOCR: thread-local declaration of 'job_t* OCR_JOB' follows non-thread-local declaration
Usage
Usage
Home
Download
Patched openbabel compilation failed on Ubuntu
Glad it was easy to resolve.
This is solved after adding #include <ctime> in bunch of locations.</ctime>
Patched openbabel compilation failed on Ubuntu
Version bump to 2.2.1
Rearranged Scaling processing
Fine-tuning scaling process
Modified Scaling processing.
Fixed thresholding for large pdfs
Separate pdf_resolution option
Fixed ocr_page for very large pages
Fixed segfault for large images morphological transform
Version bump to 2.2.0
Handle large pdf pages.
Do not try 180 degree rotation, only 90 and 270
Version bump to 2.1.12
Fix for bracket detection
Fixed confidence estimate for multi-fragment box.
Replaced string_view over rvalue to string to avoid UB.
Fixed isdigit check for AliasData
Version bump to 2.1.11
Improved handling of brackets.
Moved 2d coordinate generation before structure modifications
Reduced threshold for bracket detection.
Improved bracket recognition. Add coordinates to mis-placed superatoms.
Improved aromatic ring perception
Adjusted old-style aromatic bond selection and bracket detection.
Bug fix
Sync with customer-developed spelling/superatom maps
Added new spelling/superatom patterns
Resolution selection per box, instead of per page.
Updates for better handling metals
Added native SuspendThread on win32 build.
Version bump to 2.1.10
Added check for black lines thickness.
Implemented morphological transform for salt-and-pepper noise reduaction
Version bump to 2.1.9
Added text lines removal by dilation mask transform (inactive)
Remove large fully-connected segments
Fixed issues with timeout on OSX
Bumped version to 2.1.8
Fixed a bug with stderr
Fixes for compilation on MinGW
Fixes for compiling with clang on OSX
Put the detached threads to sleep
Introduced timeout functionality
Added v3000 option to enable V3000 MDL output
Increased lowest font height for the hi-res images
Removed hard-coded 500dpi resolution
Implemented resolution heuristics
Implemented eProcessing enum
Fixed anisotropic smoothing timer issue
Optimizations and bug fixes
Fixed a bug with MOL file output
Moved superatom/spelling map loading out of image processing loop
Added option to process multiple input files or give a folder as an input.
configure: error: GOCR API check failed; make sure you have installed libgocr-dev package or check config.log
Works now, thanks
Lyubomir, When you were building gocr did you run: make libs make all install The header should not be installed in src folder, and there certainly needs to be a library built.
I have already downloaded and built gocr and openbabel from osra.sf.net. I noticed that the lib folder is empty - not sure if that should be the case. I tried this: ./configure --with-gocr-include=/home/kotop/gocr-0.50pre-patched/src --with-gocr-lib=/home/kotop/gocr-0.50pre-patched/lib But not it says it can't find pgm2asc.h which I tripled checked that it's in the src folder. I also tried pointing it to only gocr folder (/home/kotop/gocr-0.50pre-patched) but still can't find pgm2asc.h
Lyobomir, Please download and build patched versions of gocr and openbabel available from osra.sf.net: https://sourceforge.net/projects/osra/files/gocr-patched/ https://sourceforge.net/projects/osra/files/openbabel-patched/ If osra configure script does not detect GOCR or OpenBabel headers you can use command-line options to point to the installation folders, for example: --with-gocr-include=/usr/local/include/gocr/ --with-openbabel-include=/usr/local/include/openbabel3/ Hope this helps, Igor
configure: error: GOCR API check failed; make sure you have installed libgocr-dev package or check config.log
Updated spelling and superatom dictionaries
Fixed gocr call