{"id":116403,"date":"2024-06-07T03:39:51","date_gmt":"2024-06-07T01:39:51","guid":{"rendered":"https:\/\/prohoster.info\/blog\/novosti-interneta\/reliz-sistemy-raspoznavaniya-teksta-tesseract-5-4-0"},"modified":"2024-06-07T03:39:51","modified_gmt":"2024-06-07T01:39:51","slug":"reliz-sistemy-raspoznavaniya-teksta-tesseract-5-4-0","status":"publish","type":"post","link":"https:\/\/prohoster.info\/sq\/blog\/novosti-interneta\/reliz-sistemy-raspoznavaniya-teksta-tesseract-5-4-0","title":{"rendered":"L\u00ebshimi i sistemit t\u00eb njohjes s\u00eb tekstit Tesseract 5.4.0","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"
\u00cbsht\u00eb publikuar l\u00ebshimi i sistemit t\u00eb njohjes optike t\u00eb tekstit Tesseract 5.4.0, i cili mb\u00ebshtet njohjen e karaktereve UTF-8 dhe teksteve n\u00eb m\u00eb shum\u00eb se 100 gjuh\u00eb, duke p\u00ebrfshir\u00eb rusishten, kazakishten, bjellorusishten dhe ukrainishten. Rezultati mund t\u00eb ruhet si tekst i hapur, ashtu edhe n\u00eb formate HTML (hOCR), ALTO (XML), PDF dhe TSV. Fillimisht, sistemi u krijua n\u00eb vitet 1985-1995 n\u00eb laboratorin e kompanis\u00eb Hewlett Packard, n\u00eb vitin 2005 kodi u hap n\u00ebn licenc\u00ebn Apache dhe m\u00eb von\u00eb u zhvillua me pjes\u00ebmarrjen e punonj\u00ebsve t\u00eb kompanis\u00eb Google. Tekstet burimore t\u00eb projektit shp\u00ebrndahen n\u00ebn licenc\u00ebn Apache 2.0. <\/p>\n
Tesseract p\u00ebrfshin nj\u00eb mjet konsol\u00eb dhe bibliotek\u00ebn libtesseract p\u00ebr t\u00eb inkorporuar funksionet e njohjes s\u00eb tekstit n\u00eb aplikacione t\u00eb tjera. Midis GUI-ve t\u00eb treta q\u00eb mb\u00ebshtesin Tesseract, p\u00ebrmendim gImageReader, VietOCR dhe YAGF. Oferohen dy motor\u00eb njohjeje: nj\u00eb klasik, q\u00eb njeh tekstin n\u00eb nivelin e shablloneve t\u00eb simboleve individuale, dhe nj\u00eb e re, e bazuar n\u00eb p\u00ebrdorimin e nj\u00eb sistemi t\u00eb m\u00ebsimit t\u00eb makines me rrjetin nervor rekurent LSTM, e optimizuar p\u00ebr njohjen e linjave t\u00eb plota dhe q\u00eb lejon nj\u00eb rritje t\u00eb r\u00ebnd\u00ebsishme t\u00eb sakt\u00ebsis\u00eb. Modelet e trajnuara jan\u00eb publikuar p\u00ebr 123 gjuh\u00eb. P\u00ebr optimizimin e performanc\u00ebs, ofrohen module q\u00eb p\u00ebrdorin OpenMP dhe instrukcione SIMD AVX2, AVX, AVX512F, NEON ose SSE4.1. <\/p>\n
P\u00ebrmir\u00ebsimet kryesore: <\/p>\n
\n
Shtuar mb\u00ebshtetje p\u00ebr vizatimin dhe eksportin n\u00eb formatin PAGE-XML.\n
Realizuar mund\u00ebsia e trajnimit t\u00eb modelit duke p\u00ebrdorur skedar\u00eb n\u00eb format PNG n\u00eb vend t\u00eb skedar\u00ebve LSTMF.\n
P\u00ebrmir\u00ebsuar vizatimi n\u00eb format PDF.\n
Zgjeruar API p\u00ebr p\u00ebrcaktimin e k\u00ebndit t\u00eb tekstit.\n
Zgjidhur problemet e performanc\u00ebs q\u00eb jan\u00eb identifikuar gjat\u00eb skanimit n\u00eb sistemin Coverity. <\/ul>\n