Imago OCR 2.0

22 February 2013

  • Recognition quality:
    • Added adaptive preliminary filters designed for recognition of low-quality scans or even photos;
    • Recognition rate is significantly improved due to usage of highly-adaptive methods instead of strict criterias;
    • Restriction constants and recognition parameters are moved into configurable clusters, so the better recognition rate can be achieved on various non-typical image sets using machine learning approach;
    • Machine learning tool is embedded in console application, added ability to quickly check recognition rate on specified collection;
  • Chemical features:
    • Chemical feature set is extended: bridged bonds, query features, R-groups are handled properly;
    • Superatom labels validation logic introduced which greatly increases recognition probability of complex superatom labels;
    • Abbreviation expansion support added;
  • Improved interoperability:
    • Added support of various popular image formats: JPG, BMP, DIB, TIFF, PBM, RAS and others;
    • Implemented the new logging system which produces nice human-readable reports;
    • Replaced characters recognition system, the new one supports user-defined fonts;
    • Updated API and Java / C wrappers;
    • Updated build system, reduced dependencies list;
    • Fixed a lot of bugs, overall stability is significantly improved, decreased memory consumption.

Imago OCR 1.0

28 March 2011

  • Recognizable Molecule Features
    • Single, double, triple bonds;
    • Atom labels, subscripts, isotopes, charges;
    • Superatoms;
    • Rings;
    • Stereochemistry (up- and down-bonds);
  • Supported PNG image format;
  • Implemented characters recognition system based on Fourier Descriptors;
  • Java and C wrappers introduced;
  • Ego, a Java GUI application introduced
  • Improved techniques are used for molecular graph extraction.