optical character recognition system release , supporting UTF-8 character recognition and texts in over 100 languages, including Russian, Kazakh, Belarusian, and Ukrainian. The output can be saved as plain text or in HTML (hOCR), ALTO (XML), PDF, and TSV formats. The system was originally created between 1985 and 1995 in a laboratory of Hewlett Packard, with the code being opened under the Apache license in 2005 and further developed with the participation of Google employees. The project's source texts lisensiga Apache 2.0.
Tesseract includes a console utility and the libtesseract library for embedding text recognition functions into other applications. Among the third-party GUI interfaces that support Tesseract vÔib mÀrkida , ja . Two recognition engines are offered: the classic one, recognizing text at the level of individual character patterns, and the new one based on a machine learning system using a recurrent neural network (LSTM) optimized for recognizing whole lines, allowing for a significant increase in accuracy. Pre-trained models have been released for . To optimize performance, modules utilizing OpenMP and SIMD instructions AVX2, AVX, or SSE4.1 are offered.
Peamised in Tesseract 4.1:
- The ability to output in XML format has been added (Analyzed Layout and Text Object). To use this format, run the application as «tesseract image_name output_directory alto»;
- New LSTMBox and WordStrBox rendering modules have been added, simplifying engine training;
- Support for pseudographics in hOCR output (HTML) has been added;
- Alternative Python scripts for training the machine learning-based engine have been added;
- Optimizations using AVX, AVX2, and SSE instructions have been expanded;
- By default, OpenMP support is disabled due to performance issues;
- The LSTM engine has added support for white and black lists;
- Improved build scripts based on CMake.
Allikas: opennet.ru
