Ua paʻi ʻia ka hoʻokuʻu ʻana o Tesseract 5.1 optical text recognition system, e kākoʻo ana i ka ʻike ʻana i nā huaʻōlelo UTF-8 a me nā kikokikona ma mua o 100 mau ʻōlelo, me ka Russian, Kazakh, Belarusian a me Ukrainian. Hiki ke mālama ʻia ka hopena ma ka kikokikona maʻamau a i ʻole HTML (hOCR), ALTO (XML), PDF a me nā palapala TSV. Ua hoʻokumu mua ʻia ka ʻōnaehana ma 1985-1995 i ka hale hana Hewlett Packard; ma 2005, ua wehe ʻia ke code ma lalo o ka laikini Apache a ua hoʻomohala hou ʻia me ke komo ʻana o nā limahana Google. Hāʻawi ʻia ke kumu kumu o ka papahana ma lalo o ka laikini Apache 2.0.
Loaʻa iā Tesseract kahi mea hoʻohana console a me ka waihona libtesseract no ka hoʻopili ʻana i ka hana OCR i nā noi ʻē aʻe. ʻO nā loulou GUI ʻaoʻao ʻekolu e kākoʻo ana iā Tesseract me gImageReader, VietOCR a me YAGF. Hāʻawi ʻia ʻelua ʻenekini hoʻomaopopo: ʻo kahi mea maʻamau e ʻike i ke kikokikona ma ke ʻano o nā ʻano hiʻohiʻona o kēlā me kēia kanaka, a me kahi mea hou e pili ana i ka hoʻohana ʻana i kahi ʻōnaehana aʻo mīkini e pili ana i kahi LSTM recurrent neural network, i hoʻopaʻa ʻia no ka ʻike ʻana i nā kaula holoʻokoʻa a hiki i kahi. piʻi nui i ka pololei. Ua paʻi ʻia nā kumu hoʻohālike i mākaukau no 123 mau ʻōlelo. No ka hoʻomaikaʻi ʻana i ka hana, hāʻawi ʻia nā modula e hoʻohana ana i nā kuhikuhi OpenMP a me SIMD AVX2, AVX, NEON a i ʻole SSE4.1.
Nā hoʻomaikaʻi nui ma Tesseract 5.1:
- Ua hoʻokō ʻia ka hiki ke hana i nā wahi me nā kiʻi a me nā laina ke hoʻopuka ʻia ma ALTO, hOCR a me nā ʻano kikokikona.
- Hoʻohui ʻia ka ʻāpana hou curl_timeout lkz curl_easy_setop.
- Hoʻomaikaʻi ʻia ka ʻōnaehana kūkulu.
- Ua hana ʻia e wehe i ke code i hoʻohana ʻole ʻia
- Ua hoʻopaʻa ʻia nā pōʻino ma muli o ka lawelawe hewa ʻana i nā kuhikuhi null ma ka PageIterator:: papa kuhikuhi.
Source: opennet.ru