ocr-stream 2.3.3__tar.gz → 2.4__tar.gz

{ocr_stream-2.3.3 → ocr_stream-2.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: ocr_stream
-Version: 2.3.3
+Version: 2.4
 Summary: Extrair textos de documentos digitalizados e imagens.
 Author: Bruno Chaves
 Requires-Python: >=3.11
@@ -12,7 +12,7 @@ Requires-Dist: pandas
 Requires-Dist: requests>=2.32.4
 Requires-Dist: opencv-python
 Requires-Dist: pytesseract
-Requires-Dist: convert_stream>=2.3.3
+Requires-Dist: convert_stream>=2.4
 Requires-Dist: soup_files>=1.4
 
 # ocrlib

{ocr_stream-2.3.3 → ocr_stream-2.4}/ocr_stream/__init__.py

@@ -1,38 +1,75 @@
 #!/usr/bin/env python3
 from __future__ import annotations
 from io import BytesIO
+import threading
 from ocr_stream.modules import *
 from ocr_stream.bin_tess import BinTesseract, get_path_tesseract_sys
 from ocr_stream.ocr import TesseractOcr, TextRecognized
 from ocr_stream.recognize import RecognizeImage, RecognizePdf
-from convert_stream import DocumentPdf
-from convert_stream.image import ImageObject
+from convert_stream import DocumentPdf, PageDocumentPdf
+from convert_stream.image import ImageObject, CollectionImages
 from convert_stream.mod_types import (
-    ColumnsTable, ColumnBody, HeadCell, HeadValues, create_void_table, create_void_map
+    ColumnsTable, ColumnBody, HeadCell, HeadValues, DictTextTable
 )
+
 from convert_stream.text import FindText
-from convert_stream.pdf import ConvertPdfToImages
+from convert_stream.pdf import ConvertPdfToImages, CollectionPagePdf
 import pandas as pd
 from soup_files import (
-    JsonConvert, ProgressBarAdapter, CreatePbar, Directory, File, InputFiles, LibraryDocs
+    JsonConvert, ProgressBarAdapter, ABCProgressBar, CreatePbar, Directory, File, InputFiles, LibraryDocs,
+    ProgressBarSimple
 )
 
 
+class VoidPbar(ABCProgressBar):
+
+    def __init__(self):
+        super().__init__()
+
+    def set_percent(self, percent: float):
+        pass
+
+    def set_text(self, text: str):
+        pass
+
+
+class ProgressVoid(ProgressBarAdapter):
+
+    def __init__(self, progress_bar: ABCProgressBar = VoidPbar()):
+        super().__init__(progress_bar)
+
+    def update(self, percent: float, status: str = "-"):
+        pass
+
+    def update_percent(self, percent: float = 0):
+        pass
+
+    def update_text(self, text: str = "-"):
+        pass
+
+
+def read_image(img: ImageObject, rec: RecognizeImage = RecognizeImage()) -> TextRecognized:
+    _txt = rec.image_recognize(img)
+    _txt.metadata = img.metadata
+    return _txt
+
+
 def read_images(
             images: list[ImageObject], *,
+            rec: RecognizeImage = RecognizeImage(),
             pbar: ProgressBarAdapter = CreatePbar().get()
         ) -> DocumentPdf:
     pbar.start()
     print()
     max_num = len(images)
     pages = []
-    rec = RecognizeImage()
+
     for idx, image in enumerate(images):
         pbar.update(
             ((idx + 1) / max_num) * 100,
             f'[OCR] {idx + 1} / {max_num}',
         )
-        out = rec.image_recognize(image).to_page_pdf()
+        out = read_image(image, rec=rec).to_page_pdf()
         pages.append(out)
     tmp_doc = DocumentPdf.create_from_pages(pages)
     print()
@@ -47,18 +84,31 @@ def read_document(
         ) -> DocumentPdf:
     pbar.start()
     print()
-    convert = ConvertPdfToImages.create(document)
-    convert.set_pbar(pbar)
-    images = convert.to_images(dpi=dpi)
+    _metadata = document.metadata
+    recognized_items: list[TextRecognized] = []
+    pages_pdf: list[PageDocumentPdf] = document.to_pages()
+    total = len(pages_pdf)
+    void_pbar = ProgressVoid()
+    for num, page in enumerate(pages_pdf):
+        pbar.update(
+            ((num + 1) / total) * 100,
+            f'[OCR PDF] {num + 1}/{total} {page.metadata.name}',
+        )
+        convert: ConvertPdfToImages = ConvertPdfToImages(DocumentPdf.create_from_pages([page]))
+        convert.set_pbar(void_pbar)
+        images: list[ImageObject] = convert.to_images(dpi=dpi)
+        for im in images:
+            recognized_items.append(read_image(im))
     print()
-    return read_images(images, pbar=pbar)
+    final_pages: list[PageDocumentPdf] = [x.to_page_pdf() for x in recognized_items]
+    final_doc: DocumentPdf = DocumentPdf.create_from_pages(final_pages)
+    final_doc.metadata = _metadata
+    return final_doc
 
 
 class SearchableTextImage(object):
     """
 
-    create_void_map() -> dict[str, ColumnBody]:
-
     return {
             ColumnsTable.KEY.value: ColumnBody(ColumnsTable.KEY.value, ListString([])),
             ColumnsTable.NUM_PAGE.value: ColumnBody(ColumnsTable.NUM_PAGE.value, ListString([])),
@@ -72,7 +122,7 @@ class SearchableTextImage(object):
     """
     def __init__(self, silent: bool = False):
         self.silent = silent
-        self.elements: dict[str, ColumnBody] = create_void_map()
+        self.elements: DictTextTable = DictTextTable.create_void_dict()
 
     def __repr__(self):
         return f'SearchableTextPdf: {self.elements}'
@@ -149,7 +199,7 @@ class SearchableTextImage(object):
         return pd.DataFrame.from_dict(self.elements)
 
     def to_file_json(self, file: File):
-        """Exporta os dados da busca para um arquivo .JSON"""
+        """Exporta os dados da busca para um arquivo JSON"""
         dt = JsonConvert.from_dict(self.elements).to_json_data()
         dt.to_file(file)
 
@@ -186,10 +236,10 @@ class ImageFinder(object):
         elif isinstance(img, BytesIO):
             img = ImageObject(img)
 
-        if img.metadata['file_path'] is not None:
-            self.docs_collection[img.metadata['file_path']] = img
+        if not img.metadata.file_path.is_empty:
+            self.docs_collection[img.metadata.file_path] = img
         else:
-            self.docs_collection[img.metadata['name']] = img
+            self.docs_collection[img.metadata.name] = img
 
     def add_images(self, images: list[File] | list[ImageObject] | list[bytes] | list[BytesIO]) -> None:
         for f in images:
@@ -224,7 +274,7 @@ class ImageFinder(object):
             current_img: ImageObject = self.docs_collection[file_key]
             txt_in_img = self.recognize.image_to_string(current_img)
             if (txt_in_img == '') or (txt_in_img is None):
-                self.pbar.update_text(f'ERRO: {n+1} {current_img.metadata["name"]}')
+                self.pbar.update_text(f'ERRO: {n+1} {current_img.metadata.name}')
                 continue
             try:
                 fd = FindText(txt_in_img, separator=separator)
@@ -235,20 +285,15 @@ class ImageFinder(object):
             except Exception as err:
                 print(f'{__class__.__name__} {err}')
             else:
-                file_path = current_img.metadata["file_path"]
-                file_name = current_img.metadata["name"]
-                dir_path = current_img.metadata["dir_path"]
-                file_type = current_img.metadata["extension"]
-
                 new_line: dict[str, str] = {
                     ColumnsTable.KEY.value: f'{idx}',
                     ColumnsTable.NUM_PAGE.value: 'nan',
                     ColumnsTable.NUM_LINE.value: f'{idx + 1}',
                     ColumnsTable.TEXT.value: math_text,
-                    ColumnsTable.FILE_NAME.value: file_name if file_path is not None else 'nan',
-                    ColumnsTable.FILETYPE.value: file_type if file_path is not None else 'nan',
-                    ColumnsTable.FILE_PATH.value: file_path if file_path is not None else 'nan',
-                    ColumnsTable.DIR.value: dir_path if file_path is not None else 'nan',
+                    ColumnsTable.FILE_NAME.value: current_img.metadata.file_path,
+                    ColumnsTable.FILETYPE.value: current_img.metadata.extension,
+                    ColumnsTable.FILE_PATH.value: current_img.metadata.dir_path,
+                    ColumnsTable.DIR.value: current_img.metadata.dir_path,
                 }
                 _searchable.add_line(new_line)
                 return _searchable
@@ -280,7 +325,7 @@ class ImageFinder(object):
             current_img: ImageObject = self.docs_collection[file_key]
             txt_in_img = self.recognize.image_to_string(current_img)
             if (txt_in_img == '') or (txt_in_img is None):
-                self.pbar.update_text(f'ERRO: {n + 1} {current_img.metadata["name"]}')
+                self.pbar.update_text(f'ERRO: {n + 1} {current_img.metadata.name}')
                 continue
             try:
                 fd = FindText(txt_in_img, separator=separator)
@@ -291,20 +336,15 @@ class ImageFinder(object):
             except Exception as err:
                 print(f'{__class__.__name__} {err}')
             else:
-                file_path = current_img.metadata["file_path"]
-                file_name = current_img.metadata["name"]
-                dir_path = current_img.metadata["dir_path"]
-                file_type = current_img.metadata["extension"]
-
                 new_line: dict[str, str] = {
                     ColumnsTable.KEY.value: f'{idx}',
                     ColumnsTable.NUM_PAGE.value: 'nan',
                     ColumnsTable.NUM_LINE.value: f'{idx + 1}',
                     ColumnsTable.TEXT.value: math_text,
-                    ColumnsTable.FILE_NAME.value: file_name if file_path is not None else 'nan',
-                    ColumnsTable.FILETYPE.value: file_type if file_path is not None else 'nan',
-                    ColumnsTable.FILE_PATH.value: file_path if file_path is not None else 'nan',
-                    ColumnsTable.DIR.value: dir_path if file_path is not None else 'nan',
+                    ColumnsTable.FILE_NAME.value: current_img.metadata.file_path,
+                    ColumnsTable.FILETYPE.value: current_img.metadata.extension,
+                    ColumnsTable.FILE_PATH.value: current_img.metadata.dir_path,
+                    ColumnsTable.DIR.value: current_img.metadata.dir_path,
                 }
                 _searchable.add_line(new_line)
         return _searchable

{ocr_stream-2.3.3 → ocr_stream-2.4}/ocr_stream/modules/__init__.py

@@ -8,11 +8,13 @@ class LibOcr(Enum):
 
     PYTESSERACT = "pytesseract"
     PYOCR = "pyocr"
+    EASYOCR = "easyocr"
     NOT_IMPLEMENTED = "not_implemented"
 
 
 MOD_PYTESSERACT: bool = False
 MOD_PYOCR: bool = False
+MOD_EASYOCR: bool = False
 
 try:
     from pytesseract import pytesseract, Output
@@ -27,16 +29,24 @@ try:
     import pyocr.tesseract
     MOD_PYOCR = True
 except Exception as err:
-    print(err)
+    #print(err)
     pyocr = object
-    #pyocr.tesseract = object
 
-if MOD_PYTESSERACT and MOD_PYOCR:
-    ModuleOcr: TypeAlias = Union[pyocr, pytesseract]
+
+try:
+    import easyocr
+    MOD_EASYOCR = True
+except Exception as err:
+    easyocr = object
+
+if MOD_PYTESSERACT and MOD_PYOCR and MOD_EASYOCR:
+    ModuleOcr: TypeAlias = Union[esasyocr, pytesseract, pyocr]
     DEFAULT_LIB_OCR = LibOcr.PYTESSERACT
 elif MOD_PYTESSERACT:
     ModuleOcr: TypeAlias = Union[pytesseract]
     DEFAULT_LIB_OCR = LibOcr.PYTESSERACT
+elif MOD_EASYOCR:
+    ModuleOcr: TypeAlias = Union[easyocr]
 elif MOD_PYOCR:
     ModuleOcr: TypeAlias = Union[pyocr]
     DEFAULT_LIB_OCR = LibOcr.PYOCR
@@ -44,6 +54,3 @@ else:
     ModuleOcr: TypeAlias = None
     DEFAULT_LIB_OCR = LibOcr.NOT_IMPLEMENTED
 
-
-__all__ = ['LibOcr', 'DEFAULT_LIB_OCR']
-

{ocr_stream-2.3.3 → ocr_stream-2.4}/ocr_stream/ocr/__init__.py

@@ -10,19 +10,13 @@ from soup_files import File
 from convert_stream import DocumentPdf, PageDocumentPdf
 from convert_stream.mod_types.modules import DEFAULT_LIB_IMAGE, LibImage
 from convert_stream.image import ImageObject
-from convert_stream.mod_types import get_void_metadata
+from convert_stream.mod_types import MetaDataItem, MetaDataFile
 from convert_stream.mod_types.table_types import ColumnBody
 from ocr_stream.modules import (
-    LibOcr, DEFAULT_LIB_OCR, pytesseract, Output, pyocr,
+    LibOcr, DEFAULT_LIB_OCR, pytesseract, Output, pyocr, easyocr
 )
 from ocr_stream.bin_tess import BinTesseract
 
-try:
-    import pyocr.tesseract
-except Exception as err:
-    print(err)
-    pyocr = None
-
 
 class TextRecognized(object):
     """
@@ -31,8 +25,9 @@ class TextRecognized(object):
     """
 
     def __init__(self, bytes_recognized: bytes):
-        self.metadata: dict[str, str | None] = get_void_metadata()
+        self.metadata: MetaDataFile = MetaDataFile()
         self.bytes_recognized: bytes = bytes_recognized
+
         self.list_bad_char: list[str] = [
             ':', ',', ';', '$', '=',
             '!', '}', '{', '(', ')',
@@ -51,7 +46,10 @@ class TextRecognized(object):
         return doc.to_pages()[0]
 
     def to_document(self) -> DocumentPdf:
-        return DocumentPdf(BytesIO(self.bytes_recognized))
+        tmp_doc = DocumentPdf(BytesIO(self.bytes_recognized))
+        tmp_doc.metadata.file_path = self.metadata.file_path
+        tmp_doc.metadata.dir_path = self.metadata.dir_path
+        return tmp_doc
 
     def to_dict(self, separator: str = '\n') -> dict[str, ColumnBody]:
         return self.to_page_pdf().to_dict(separator=separator)
@@ -138,6 +136,78 @@ class IPytesseract(ABCOcrTesseract):
         return TextRecognized(bt)
 
 
+# ======================================================================#
+# Modulo easyocr
+# ======================================================================#
+class IEasyOcr(ABCOcrTesseract):
+
+    def __init__(self, bin_tess: BinTesseract, *, lang: str = 'pt', tess_data_dir: str = None):
+        super().__init__(bin_tess, lang=lang, tess_data_dir=tess_data_dir)
+        self.reader: easyocr.Reader = easyocr.Reader([self.lang], gpu=True)
+
+    def to_string(self, img: Union[File, ImageObject]) -> str:
+        if isinstance(img, File):
+            img = ImageObject(img)
+        result = self.reader.readtext(img.to_opencv())
+        text = '\n'.join([res[1] for res in result])
+        return text
+
+    def to_reconize(self, img: Union[File, ImageObject]) -> TextRecognized:
+        import cv2
+        import numpy as np
+        from PIL import Image
+        from io import BytesIO
+        from reportlab.pdfgen import canvas
+        from reportlab.lib.pagesizes import A4
+        from reportlab.lib.utils import ImageReader
+
+        # 1️⃣ Converter o objeto de entrada para imagem OpenCV
+        if isinstance(img, File):
+            img = ImageObject(img)
+        img_cv = img.to_opencv()
+        img_h, img_w = img_cv.shape[:2]
+
+        # 2️⃣ Reconhecer o texto com EasyOCR
+        results = self.reader.readtext(img_cv)
+
+        # 3️⃣ Converter a imagem para PDF (imagem de fundo)
+        buffer = BytesIO()
+        pdf = canvas.Canvas(buffer, pagesize=(img_w, img_h))
+
+        # Desenha a imagem original como fundo
+        img_pil = Image.fromarray(cv2.cvtColor(img_cv, cv2.COLOR_BGR2RGB))
+        pdf.drawImage(ImageReader(img_pil), 0, 0, width=img_w, height=img_h)
+
+        # 4️⃣ Adiciona o texto OCR como camada "invisível" sobre a imagem
+        pdf.setFont("Helvetica", 8)
+        pdf.setFillColorRGB(1, 1, 1, alpha=0)  # texto invisível
+
+        for (bbox, text, conf) in results:
+            if not text.strip():
+                continue
+            # Coordenadas do bounding box (EasyOCR retorna 4 pontos)
+            (x1, y1), (x2, y2), (x3, y3), (x4, y4) = bbox
+
+            # Média da posição vertical
+            avg_y = (y1 + y2 + y3 + y4) / 4
+
+            # Posição invertida (PDF tem origem no canto inferior esquerdo)
+            y_pdf = img_h - avg_y
+
+            # Largura estimada
+            text_width = abs(x2 - x1)
+            pdf.drawString(x1, y_pdf, text)
+
+        pdf.showPage()
+        pdf.save()
+        buffer.seek(0)
+
+        # 5️⃣ Retorna como TextRecognized (bytes do PDF com texto embutido)
+        text_rec = TextRecognized(buffer.read())
+        text_rec.metadata = img.metadata if hasattr(img, "metadata") else MetaDataFile()
+        return text_rec
+
+
 # ======================================================================#
 # Modulo OCR pyocr
 # ======================================================================#
@@ -202,6 +272,10 @@ class TesseractOcr(object):
             md = IPytesseract(bin_tess, lang=lang, tess_data_dir=tess_data_dir)
         elif lib_ocr == LibOcr.PYOCR:
             md = IPyOcr(bin_tess, lang=lang, tess_data_dir=tess_data_dir)
+        elif lib_ocr == LibOcr.EASYOCR:
+            if lang is None:
+                lang = 'pt'
+            md = IEasyOcr(bin_tess, lang=lang, tess_data_dir=tess_data_dir)
         else:
             raise ValueError(f'{lib_ocr} is not supported')
         return cls(md)

{ocr_stream-2.3.3 → ocr_stream-2.4}/ocr_stream/recognize.py

@@ -123,4 +123,3 @@ class RecognizePdf(object):
         recognized_doc = self.recognize_document(file_doc, dpi=dpi)
         del file_doc
         return recognized_doc
-

{ocr_stream-2.3.3 → ocr_stream-2.4}/ocr_stream.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: ocr_stream
-Version: 2.3.3
+Version: 2.4
 Summary: Extrair textos de documentos digitalizados e imagens.
 Author: Bruno Chaves
 Requires-Python: >=3.11
@@ -12,7 +12,7 @@ Requires-Dist: pandas
 Requires-Dist: requests>=2.32.4
 Requires-Dist: opencv-python
 Requires-Dist: pytesseract
-Requires-Dist: convert_stream>=2.3.3
+Requires-Dist: convert_stream>=2.4
 Requires-Dist: soup_files>=1.4
 
 # ocrlib

{ocr_stream-2.3.3 → ocr_stream-2.4}/ocr_stream.egg-info/requires.txt

@@ -5,5 +5,5 @@ pandas
 requests>=2.32.4
 opencv-python
 pytesseract
-convert_stream>=2.3.3
+convert_stream>=2.4
 soup_files>=1.4

{ocr_stream-2.3.3 → ocr_stream-2.4}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "ocr_stream"
-version = "2.3.3"
+version = "2.4"
 authors = [
   { name="Bruno Chaves"},
 ]
@@ -21,7 +21,7 @@ dependencies = [
     "requests>=2.32.4",
     "opencv-python",
     "pytesseract",
-    "convert_stream>=2.3.3",
+    "convert_stream>=2.4",
     "soup_files>=1.4",
 ]