Programe native: demonstrații în terminal

LegatorOCR transformă pagini scanate într-un PDF cu text. Aici sunt trei imagini cu text ușor înclinat.

• $ ls -l page-*.png

  -rw-r--r-- 1 root root 179783 Nov  3 14:20 page-1.png
  -rw-r--r-- 1 root root 175232 Nov  3 14:20 page-2.png
  -rw-r--r-- 1 root root 187296 Nov  3 14:20 page-3.png
  

O singură comandă corectează înclinarea, recunoaște textul cu OCR și scrie PDF-ul și un fișier text. Tabelul de timpuri arată unde se duce timpul.

• $ legatorocr page-*.png --format none --pdf book.pdf --ocr --lang en --txt book.txt --timings

  OCR engine: tesseract 5.3.4, language: en
  page-1: 788x1491 px @ 150 dpi, skew +2.33°
  page-2: 789x1491 px @ 150 dpi, skew -1.65°
  page-3: 788x1491 px @ 150 dpi, skew +3.10°
  stage                  calls     time (s)   share   per call (ms)
  ocr                          3         5.84   65.6%         1946.2
  clean-up                     3         3.02   34.0%         1007.5
    deskew/rotate              3         1.04   11.7%          345.8
    deskew/pictures            3         0.26    2.9%           87.3
    deskew/detect              3         0.24    2.7%           79.7
    content/analyse            3         0.18    2.1%           61.3
    deskew/analyse             3         0.17    1.9%           55.8
    content/detect             3         0.04    0.5%           14.5
  load                         3         0.04    0.5%           13.4
    content/grayscale          3         0.01    0.2%            4.8
    deskew/grayscale           3         0.01    0.2%            4.8
    clean-up/background        3         0.01    0.1%            2.5
  orientation                  3         0.00    0.0%            0.0
  total (summed over threads): 8.90 s
  peak memory: 203 MiB
  wrote book.pdf (3 page(s), 155 OCR words)
  wrote book.txt
  

Textul recunoscut, pagină cu pagină. Imaginile din demonstrație sunt generate, deci rezultatul e curat; scanările reale sunt mai grele.

• $ ls -l book.pdf book.txt

  -rw-r--r-- 1 root root 242042 Nov  3 14:20 book.pdf
  -rw-r--r-- 1 root root    839 Nov  3 14:20 book.txt
  

• $ head -n 8 book.txt

  Chapter 1. The Beginnings
  The library was founded in 1864 by a
  group of teachers and merchants. In the
  first years the books were kept in a single
  room, and lending was recorded in a ledger.
  By the end of the century the collection
  held over ten thousand volumes, mostly gifts.
  Chapter 2. The Beginnings
  

PDF-ul are strat de text: poate fi căutat și copiat. Programul rulează local, fără cont.

• $ pdftotext book.pdf - | head -n 3

  Chapter 1. The Beginnings
  The library was founded in 1864 by a
  group of teachers and merchants. In the