API tools faq
paste
zdenop

Simple example of python wrapper for tesseract and leptonica

zdenop
Dec 25th, 2020 (edited)
1,597
0
Never
Add comment
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
Python 10.44 KB | None | 0 0
raw download clone embed print report
  1. #!/usr/bin/env python3
  2. # -*- coding: utf-8 -*-
  3.  
  4. """Simple example of python wrapper for tesseract and leptonica
  5.   based on:
  6.   https://sk-spell.sk.cx/building-minimalistic-tesseract
  7.   https://github.com/zdenop/SimpleTesseractPythonWrapper
  8. """
  9.  
  10. import ctypes
  11. import locale
  12. import os
  13. import platform
  14. from ctypes.util import find_library
  15.  
  16. import cffi
  17. from PIL import Image, ImageDraw, ImageFont
  18.  
  19. ffi = cffi.FFI()
  20. ffi.cdef(
  21.     """
  22. typedef signed char             l_int8;
  23. typedef unsigned char           l_uint8;
  24. typedef short                   l_int16;
  25. typedef unsigned short          l_uint16;
  26. typedef int                     l_int32;
  27. typedef unsigned int            l_uint32;
  28. typedef float                   l_float32;
  29. typedef double                  l_float64;
  30. typedef long long               l_int64;
  31. typedef unsigned long long      l_uint64;
  32. typedef int l_ok; /*!< return type 0 if OK, 1 on error */
  33.  
  34.  
  35. struct Pix;
  36. typedef struct Pix PIX;
  37. typedef enum lept_img_format {
  38.    IFF_UNKNOWN        = 0,
  39.    IFF_BMP            = 1,
  40.    IFF_JFIF_JPEG      = 2,
  41.    IFF_PNG            = 3,
  42.    IFF_TIFF           = 4,
  43.    IFF_TIFF_PACKBITS  = 5,
  44.    IFF_TIFF_RLE       = 6,
  45.    IFF_TIFF_G3        = 7,
  46.    IFF_TIFF_G4        = 8,
  47.    IFF_TIFF_LZW       = 9,
  48.    IFF_TIFF_ZIP       = 10,
  49.    IFF_PNM            = 11,
  50.    IFF_PS             = 12,
  51.    IFF_GIF            = 13,
  52.    IFF_JP2            = 14,
  53.    IFF_WEBP           = 15,
  54.    IFF_LPDF           = 16,
  55.    IFF_TIFF_JPEG      = 17,
  56.    IFF_DEFAULT        = 18,
  57.    IFF_SPIX           = 19
  58. };
  59.  
  60. typedef enum newsev {
  61.    L_SEVERITY_EXTERNAL = 0,   /* Get the severity from the environment   */
  62.    L_SEVERITY_ALL      = 1,   /* Lowest severity: print all messages     */
  63.    L_SEVERITY_DEBUG    = 2,   /* Print debugging and higher messages     */
  64.    L_SEVERITY_INFO     = 3,   /* Print informational and higher messages */
  65.    L_SEVERITY_WARNING  = 4,   /* Print warning and higher messages       */
  66.    L_SEVERITY_ERROR    = 5,   /* Print error and higher messages         */
  67.    L_SEVERITY_NONE     = 6    /* Highest severity: print no messages     */
  68. };
  69.  
  70. char * getLeptonicaVersion (  );
  71. PIX * pixRead ( const char *filename );
  72. PIX * pixCreate ( int width, int height, int depth );
  73. PIX * pixEndianByteSwapNew(PIX  *pixs);
  74. l_int32 pixSetData ( PIX *pix, l_uint32 *data );
  75. l_ok pixSetPixel ( PIX *pix, l_int32 x, l_int32 y, l_uint32 val );
  76. l_ok pixWrite ( const char *fname, PIX *pix, l_int32 format );
  77. l_int32 pixFindSkew ( PIX *pixs, l_float32 *pangle, l_float32 *pconf );
  78. PIX * pixDeskew ( PIX *pixs, l_int32 redsearch );
  79. void pixDestroy ( PIX **ppix );
  80. l_ok pixGetResolution ( const PIX *pix, l_int32 *pxres, l_int32 *pyres );
  81. l_ok pixSetResolution ( PIX *pix, l_int32 xres, l_int32 yres );
  82. l_int32 pixGetWidth ( const PIX *pix );
  83. l_int32 setMsgSeverity ( l_int32 newsev );
  84.  
  85. typedef struct TessBaseAPI TessBaseAPI;
  86. typedef struct ETEXT_DESC ETEXT_DESC;
  87. typedef struct TessPageIterator TessPageIterator;
  88. typedef struct TessResultIterator TessResultIterator;
  89. typedef int BOOL;
  90.  
  91. typedef enum TessOcrEngineMode  {
  92.    OEM_TESSERACT_ONLY          = 0,
  93.    OEM_LSTM_ONLY               = 1,
  94.    OEM_TESSERACT_LSTM_COMBINED = 2,
  95.    OEM_DEFAULT                 = 3} TessOcrEngineMode;
  96.  
  97. typedef enum TessPageSegMode {
  98.    PSM_OSD_ONLY               =  0,
  99.    PSM_AUTO_OSD               =  1,
  100.    PSM_AUTO_ONLY              =  2,
  101.    PSM_AUTO                   =  3,
  102.    PSM_SINGLE_COLUMN          =  4,
  103.    PSM_SINGLE_BLOCK_VERT_TEXT =  5,
  104.    PSM_SINGLE_BLOCK           =  6,
  105.    PSM_SINGLE_LINE            =  7,
  106.    PSM_SINGLE_WORD            =  8,
  107.    PSM_CIRCLE_WORD            =  9,
  108.    PSM_SINGLE_CHAR            = 10,
  109.    PSM_SPARSE_TEXT            = 11,
  110.    PSM_SPARSE_TEXT_OSD        = 12,
  111.    PSM_COUNT                  = 13} TessPageSegMode;
  112.  
  113. typedef enum TessPageIteratorLevel {
  114.     RIL_BLOCK    = 0,
  115.     RIL_PARA     = 1,
  116.     RIL_TEXTLINE = 2,
  117.     RIL_WORD     = 3,
  118.     RIL_SYMBOL    = 4} TessPageIteratorLevel;    
  119.  
  120. TessPageIterator* TessBaseAPIAnalyseLayout(TessBaseAPI* handle);
  121. TessPageIterator* TessResultIteratorGetPageIterator(TessResultIterator* handle);
  122.  
  123. BOOL TessPageIteratorNext(TessPageIterator* handle, TessPageIteratorLevel level);
  124. BOOL TessPageIteratorBoundingBox(const TessPageIterator* handle, TessPageIteratorLevel level,
  125.                                 int* left, int* top, int* right, int* bottom);
  126.  
  127. const char* TessVersion();
  128.  
  129. TessBaseAPI* TessBaseAPICreate();
  130. int    TessBaseAPIInit3(TessBaseAPI* handle, const char* datapath, const char* language);
  131. int    TessBaseAPIInit2(TessBaseAPI* handle, const char* datapath, const char* language, TessOcrEngineMode oem);
  132. void   TessBaseAPISetPageSegMode(TessBaseAPI* handle, TessPageSegMode mode);
  133. void   TessBaseAPISetImage(TessBaseAPI* handle,
  134.                           const unsigned char* imagedata, int width, int height,
  135.                           int bytes_per_pixel, int bytes_per_line);
  136. void   TessBaseAPISetImage2(TessBaseAPI* handle, struct Pix* pix);
  137.  
  138. BOOL   TessBaseAPISetVariable(TessBaseAPI* handle, const char* name, const char* value);
  139. BOOL   TessBaseAPIDetectOrientationScript(TessBaseAPI* handle, char** best_script_name,
  140.                                                            int* best_orientation_deg, float* script_confidence,
  141.                                                            float* orientation_confidence);
  142. int TessBaseAPIRecognize(TessBaseAPI* handle, ETEXT_DESC* monitor);
  143. TessResultIterator* TessBaseAPIGetIterator(TessBaseAPI* handle);
  144. BOOL   TessResultIteratorNext(TessResultIterator* handle, TessPageIteratorLevel level);
  145. char*  TessResultIteratorGetUTF8Text(const TessResultIterator* handle, TessPageIteratorLevel level);
  146. float  TessResultIteratorConfidence(const TessResultIterator* handle, TessPageIteratorLevel level);
  147. char*  TessBaseAPIGetUTF8Text(TessBaseAPI* handle);
  148. const char*  TessResultIteratorWordFontAttributes(const TessResultIterator* handle, BOOL* is_bold, BOOL* is_italic,
  149.                                                              BOOL* is_underlined, BOOL* is_monospace, BOOL* is_serif,
  150.                                                              BOOL* is_smallcaps, int* pointsize, int* font_id);
  151. void   TessBaseAPIEnd(TessBaseAPI* handle);
  152. void   TessBaseAPIDelete(TessBaseAPI* handle);
  153. """
  154. )
  155.  
  156.  
  157. def get_abs_path_of_library(library):
  158.     """Get absolute path of library."""
  159.     abs_path = None
  160.     lib_name = find_library(library)
  161.     if os.path.exists(lib_name):
  162.         abs_path = os.path.abspath(lib_name)
  163.         return abs_path
  164.     libdl = ctypes.CDLL(lib_name)
  165.     if not libdl:
  166.         return abs_path  # None
  167.     try:
  168.         dlinfo = libdl.dlinfos
  169.     except AttributeError as err:
  170.         # Workaroung for linux
  171.         abs_path = str(err).split(":")[0]
  172.     return abs_path
  173.  
  174.  
  175. def pil2PIX32(im, leptonica):
  176.     """Convert PIL to leptonica PIX."""
  177.     # At the moment we handle everything as RGBA image
  178.     if im.mode != "RGBA":
  179.         im = im.convert("RGBA")
  180.     depth = 32
  181.     width, height = im.size
  182.     data = im.tobytes("raw", "RGBA")
  183.     pixs = leptonica.pixCreate(width, height, depth)
  184.     leptonica.pixSetData(pixs, ffi.from_buffer("l_uint32[]", data))
  185.  
  186.     try:
  187.         resolutionX = im.info["resolution"][0]
  188.         resolutionY = im.info["resolution"][1]
  189.         leptonica.pixSetResolution(pixs, resolutionX, resolutionY)
  190.     except KeyError:
  191.         pass
  192.     try:
  193.         resolutionX = im.info["dpi"][0]
  194.         resolutionY = im.info["dpi"][1]
  195.         leptonica.pixSetResolution(pixs, resolutionX, resolutionY)
  196.     except KeyError:
  197.         pass
  198.  
  199.     return leptonica.pixEndianByteSwapNew(pixs)
  200.  
  201.  
  202. def img_lepto_to_pil(pix):
  203.     """Convert leptonica pix to PIL
  204.       Source: https://stackoverflow.com/questions/55195932/typeerror-initializer-for-ctype-unsigned-int-must-be-a-cdata-pointer-not-b/57776268#57776268
  205.    """
  206.     cdata_ptr = ffi.new("l_uint8**")
  207.     size_ptr = ffi.new("size_t*")
  208.     leptonica.pixWriteMem(cdata_ptr, size_ptr, pix, IFF_TIFF)
  209.     cdata = cdata_ptr[0]
  210.     size = size_ptr[0]
  211.  
  212.     tiff_bytes = bytes(ffi.buffer(cdata, size))
  213.     with BytesIO(tiff_bytes) as bytesio:
  214.         pilimage = PIL.Image.open(bytesio).copy()
  215.         return pilimag
  216.  
  217.  
  218. def main():
  219.     """Main loop."""
  220.     # Settings
  221.     tess_libname = r"F:/win64_msvc_min/bin/tesseract50.dll"
  222.     lept_libname = r"F:/win64_msvc_min/bin/leptonica-1.81.0.dll"
  223.     filename = "line.ppm"
  224.     lang = "eng"
  225.  
  226.     tessdata = os.environ.get("TESSDATA_PREFIX")
  227.     if not tessdata:
  228.         # Use project tessdata
  229.         tessdata = os.path.join(os.getcwd(), "tessdata")
  230.         os.environ["TESSDATA_PREFIX"] = tessdata
  231.  
  232.     # Load libraries in ABI mode
  233.     if not os.path.dirname(tess_libname) in os.environ.get("PATH"):
  234.         os.environ["PATH"] = os.environ.get("PATH") + ';' + os.path.dirname(tess_libname)
  235.     if os.path.exists(tess_libname):
  236.         tesseract = ffi.dlopen(tess_libname)
  237.     else:
  238.         print(f"'{tess_libname}' does not exists!")
  239.     tesseract_version = ffi.string(tesseract.TessVersion())
  240.     print("Tesseract-ocr version", tesseract_version.decode("utf-8"))
  241.  
  242.     if os.path.exists(lept_libname):
  243.         leptonica = ffi.dlopen(lept_libname)
  244.     else:
  245.         print(f"'{lept_libname}' does not exists!")
  246.     leptonica_version = ffi.string(leptonica.getLeptonicaVersion())
  247.     print(leptonica_version.decode("utf-8"))
  248.     api = None
  249.  
  250.     # Read image to pix
  251.     im = Image.open(filename)
  252.     pix = pil2PIX32(im, leptonica)
  253.  
  254.     # Turn off leptonica warnings
  255.     leptonica.setMsgSeverity(leptonica.L_SEVERITY_EXTERNAL)
  256.  
  257.     # Create tesseract API
  258.     if api:
  259.         tesseract.TessBaseAPIEnd(api)
  260.         tesseract.TessBaseAPIDelete(api)
  261.     api = tesseract.TessBaseAPICreate()
  262.  
  263.     oem = tesseract.OEM_DEFAULT
  264.     tesseract.TessBaseAPIInit2(api, tessdata.encode(), lang.encode(), oem)
  265.     tesseract.TessBaseAPISetPageSegMode(api, tesseract.PSM_AUTO)
  266.  
  267.     tesseract.TessBaseAPISetImage2(api, pix)
  268.     # recognize is needed to get result iterator
  269.     tesseract.TessBaseAPIRecognize(api, ffi.NULL)
  270.  
  271.     utf8_text = ffi.string(tesseract.TessBaseAPIGetUTF8Text(api)).decode("utf-8")
  272.     print(utf8_text)
  273.  
  274.     # Delete api and pix
  275.     if api:
  276.         tesseract.TessBaseAPIEnd(api)
  277.         tesseract.TessBaseAPIDelete(api)
  278.  
  279.     result = ffi.new("PIX**")
  280.     result[0] = pix
  281.     leptonica.pixDestroy(result)
  282.     del pix
  283.     del result
  284.     api = None
  285.  
  286.  
  287. if __name__ == "__main__":
  288.     main()
  289.  
Add Comment
Please, Sign In to add comment
Public Pastes
We use cookies for various purposes including analytics. By continuing to use Pastebin, you agree to our use of cookies as described in the Cookies Policy.  OK, I Understand
Not a member of Pastebin yet?
Sign Up, it unlocks many cool features!