Merge branch 'main' into improvement/remove_torch

CHANGELOG.md

@@ -11,6 +11,11 @@ The version numbers are according to [Semantic Versioning](http://semver.org/).
 
 ### Removed
 
+
+## Release v0.2.3 (2024-10-29)
+### Changed
+- OCR provider specific dependencies are now being loaded lazily to reduce the import time
+
 ## Release v0.2.2 (2024-10-03)
 ### Added
 - Added OpenTelemetry to `GoogleAzureOCR`, `GoogleOCR`, `AzureOCR`, and `OcrWrapper` to enable tracing of the OCR process

ocr_wrapper/aws.py

@@ -6,26 +6,8 @@
 from .bbox import BBox
 from .ocr_wrapper import OcrCacheDisabled, OcrWrapper
 
-try:
-    import boto3
-except ImportError:
-    _has_boto3 = False
-else:
-    _has_boto3 = True
-
-
-def requires_boto(fn):
-    @functools.wraps(fn)
-    def wrapper_decocator(*args, **kwargs):
-        if not _has_boto3:
-            raise ImportError('AWS Textract requires missing "boto3" package.')
-        return fn(*args, **kwargs)
-
-    return wrapper_decocator
-
 
 class AwsOCR(OcrWrapper):
-    @requires_boto
     def __init__(
         self,
         *,
@@ -34,10 +16,19 @@ def __init__(
         add_checkboxes: bool = False,
         verbose: bool = False
     ):
-        super().__init__(cache_file=cache_file, max_size=max_size, add_checkboxes=add_checkboxes, verbose=verbose)
+        try:
+            import boto3
+        except ImportError:
+            raise ImportError('AwsOCR requires missing "boto3" package.')
+
+        super().__init__(
+            cache_file=cache_file,
+            max_size=max_size,
+            add_checkboxes=add_checkboxes,
+            verbose=verbose,
+        )
         self.client = boto3.client("textract", region_name="eu-central-1")
 
-    @requires_boto
     def _get_ocr_response(self, img: Image.Image):
         """Gets the OCR response from AWS. Uses cached response if a cache file has been specified and the
         document has been OCRed already"""
@@ -52,7 +43,6 @@ def _get_ocr_response(self, img: Image.Image):
             self._put_on_shelf(img, response)
         return response
 
-    @requires_boto
     def _convert_ocr_response(self, response) -> tuple[List[BBox], dict[str, Any]]:
         """Converts the response given by Google OCR to a list of BBox"""
         bboxes = []

ocr_wrapper/azure.py

@@ -13,33 +13,9 @@
 from .ocr_wrapper import OcrCacheDisabled, OcrWrapper, Union
 from .utils import set_image_attributes
 
-try:
-    from msrest.authentication import CognitiveServicesCredentials
-    from msrest.exceptions import ClientRequestError
-
-    from azure.cognitiveservices.vision.computervision import ComputerVisionClient
-    from azure.cognitiveservices.vision.computervision.models import (
-        ComputerVisionOcrErrorException,
-        OperationStatusCodes,
-    )
-except ImportError:
-    _has_azure = False
-else:
-    _has_azure = True
-
 tracer = trace.get_tracer(__name__)
 
 
-def requires_azure(fn):
-    @functools.wraps(fn)
-    def wrapper_decocator(*args, **kwargs):
-        if not _has_azure:
-            raise ImportError('Azure Read requires missing "azure-cognitiveservices-vision-computervision" package.')
-        return fn(*args, **kwargs)
-
-    return wrapper_decocator
-
-
 def _discretize_angle_to_90_deg(rotation: float) -> int:
     """Discretize an angle to the nearest 90 degrees"""
     return int(((rotation + 45) // 90 * 90) % 360)
@@ -70,7 +46,6 @@ def _determine_endpoint_and_key(endpoint: Optional[str], key: Optional[str]) ->
 
 
 class AzureOCR(OcrWrapper):
-    @requires_azure
     def __init__(
         self,
         *,
@@ -85,6 +60,11 @@ def __init__(
         add_qr_barcodes: bool = False,
         verbose: bool = False,
     ):
+        try:
+            from msrest.authentication import CognitiveServicesCredentials
+            from azure.cognitiveservices.vision.computervision import ComputerVisionClient
+        except ImportError:
+            raise ImportError('AzureOCR requires missing "azure-cognitiveservices-vision-computervision" package.')
         super().__init__(
             cache_file=cache_file,
             max_size=max_size,
@@ -99,11 +79,19 @@ def __init__(
         endpoint, key = _determine_endpoint_and_key(endpoint, key)
         self.client = ComputerVisionClient(endpoint, CognitiveServicesCredentials(key))
 
-    @requires_azure
     @tracer.start_as_current_span(name="AzureOCR.get_ocr_response")
     def _get_ocr_response(self, img: Image.Image):
         """Gets the OCR response from the Azure. Uses cached response if a cache file has been specified and the
         document has been OCRed already"""
+        try:
+            from msrest.exceptions import ClientRequestError
+            from azure.cognitiveservices.vision.computervision.models import (
+                ComputerVisionOcrErrorException,
+                OperationStatusCodes,
+            )
+        except ImportError:
+            raise ImportError('AzureOCR requires missing "azure-cognitiveservices-vision-computervision" package.')
+
         span = trace.get_current_span()
         set_image_attributes(span, img)
 
@@ -158,7 +146,6 @@ def _get_ocr_response(self, img: Image.Image):
             self._put_on_shelf(img, read_result)
         return read_result
 
-    @requires_azure
     @tracer.start_as_current_span(name="AzureOCR._convert_ocr_response")
     def _convert_ocr_response(self, response) -> tuple[List[BBox], dict[str, Any]]:
         """Converts the response given by Azure Read to a list of BBox"""

ocr_wrapper/easy_ocr.py

@@ -9,26 +9,8 @@
 from .bbox import BBox
 from .ocr_wrapper import OcrWrapper, OcrCacheDisabled
 
-try:
-    import easyocr
-except ImportError:
-    _has_easyocr = False
-else:
-    _has_easyocr = True
-
-
-def requires_easyocr(fn):
-    @functools.wraps(fn)
-    def wrapper_decocator(*args, **kwargs):
-        if not _has_easyocr:
-            raise ImportError('Easy OCR requires missing "easyocr" package.')
-        return fn(*args, **kwargs)
-
-    return wrapper_decocator
-
 
 class EasyOCR(OcrWrapper):
-    @requires_easyocr
     def __init__(
         self,
         *,
@@ -46,6 +28,11 @@ def __init__(
             languages: A string or a list of languages to use for OCR from the list here: https://www.jaided.ai/easyocr/
             width_thr: Distance where bounding boxes are still getting merged into one
         """
+        try:
+            import easyocr
+        except ImportError:
+            raise ImportError('EasyOCR requires missing "easyocr" package.')
+
         super().__init__(
             cache_file=cache_file,
             max_size=max_size,
@@ -58,7 +45,6 @@ def __init__(
 
         self.client = easyocr.Reader(self.languages, **kwargs)
 
-    @requires_easyocr
     def _get_ocr_response(self, img: Image.Image):
         """Gets the OCR response from EasyOCR. Uses a cached response if a cache file has been specified and the
         document has been OCRed already"""
@@ -69,7 +55,6 @@ def _get_ocr_response(self, img: Image.Image):
             self._put_on_shelf(img, response)
         return response
 
-    @requires_easyocr
     def _convert_ocr_response(self, response) -> tuple[List[BBox], dict[str, Any]]:
         """Converts the response given by EasyOCR to a list of BBox"""
         bboxes, confidences = [], []

ocr_wrapper/google_azure_ocr.py

@@ -295,7 +295,6 @@ def __init__(self, bboxes: list[BBox]):
         for i, bbox in enumerate(bboxes):
             self.rtree.insert(i, bbox.get_shapely_polygon().bounds)
 
-    @tracer.start_as_current_span(name="BBoxOverlapChecker.get_overlapping_bboxes")
     def get_overlapping_bboxes(self, bbox: BBox, threshold: float = 0.01) -> list[BBox]:
         """Returns the bboxes that overlap with the given bbox.
 
@@ -318,7 +317,6 @@ def get_overlapping_bboxes(self, bbox: BBox, threshold: float = 0.01) -> list[BB
             ):
                 overlapping_bboxes.append(self.bboxes[i])
 
-        span.set_attribute("nb_overlapping_bboxes", len(overlapping_bboxes))
         return overlapping_bboxes
 
 

ocr_wrapper/google_document_ocr_checkbox_detector.py

@@ -1,8 +1,7 @@
 from __future__ import annotations
 
-import functools
 import os
-from typing import Optional, Union
+from typing import TYPE_CHECKING, Optional, Union
 
 from PIL import Image
 
@@ -11,25 +10,9 @@
 
 from .utils import resize_image
 
-try:
-    from google.api_core.client_options import ClientOptions
+# Load the Google Cloud Document AI client library globally only for type checking (needed for argument types)
+if TYPE_CHECKING:
     from google.cloud import documentai
-except ImportError:
-    _has_gcloud = False
-else:
-    _has_gcloud = True
-
-
-def requires_gcloud(fn):
-    @functools.wraps(fn)
-    def wrapper_decocator(*args, **kwargs):
-        if not _has_gcloud:
-            raise ImportError(
-                "GoogleDocumentOcrCheckboxDetector OCR requires missing 'google-cloud-documentai' package."
-            )
-        return fn(*args, **kwargs)
-
-    return wrapper_decocator
 
 
 def _val_or_env(val: Optional[str], env: str, default: Optional[str] = None) -> Optional[str]:
@@ -62,7 +45,6 @@ def _visual_element_to_bbox(visual_element) -> tuple[BBox, float]:
 
 
 class GoogleDocumentOcrCheckboxDetector:
-    @requires_gcloud
     def __init__(
         self,
         project_id: Optional[str] = None,
@@ -71,6 +53,12 @@ def __init__(
         processor_version: Optional[str] = None,
         max_size: Optional[int] = 4096,
     ):
+        try:
+            from google.api_core.client_options import ClientOptions
+            from google.cloud import documentai
+        except ImportError:
+            raise ImportError("GoogleDocumentOcrCheckboxDetector requires missing 'google-cloud-documentai' package.")
+
         self.project_id = _val_or_env(project_id, "GOOGLE_DOC_OCR_PROJECT_ID")
         self.location = _val_or_env(location, "GOOGLE_DOC_OCR_LOCATION", default="eu")
         self.processor_id = _val_or_env(processor_id, "GOOGLE_DOC_OCR_PROCESSOR_ID")
@@ -105,8 +93,12 @@ def __init__(
             self.project_id, self.location, self.processor_id, self.processor_version
         )
 
-    @requires_gcloud
     def detect_checkboxes(self, page: Union[Image.Image, documentai.RawDocument]) -> tuple[list[BBox], list[float]]:
+        try:
+            from google.cloud import documentai
+        except ImportError:
+            raise ImportError("GoogleDocumentOcrCheckboxDetector requires missing 'google-cloud-documentai' package.")
+
         if isinstance(page, Image.Image):
             if self.max_size is not None:
                 page = resize_image(img=page, max_size=self.max_size)

ocr_wrapper/google_ocr.py

@@ -1,6 +1,5 @@
 from __future__ import annotations
 
-import functools
 import os
 import time
 from time import sleep
@@ -15,24 +14,6 @@
 
 tracer = trace.get_tracer(__name__)
 
-try:
-    from google.cloud import vision
-except ImportError:
-    _has_gcloud = False
-else:
-    _has_gcloud = True
-
-
-def requires_gcloud(fn):
-    @functools.wraps(fn)
-    def wrapper_decocator(*args, **kwargs):
-        if not _has_gcloud:
-            raise ImportError('Google OCR requires missing "google-cloud-vision" package.')
-        return fn(*args, **kwargs)
-
-    return wrapper_decocator
-
-
 # Define a list of languages which are written from right to left. This is needed to determine the rotation of the document
 rtl_languages = [
     "ar",
@@ -183,7 +164,6 @@ class GoogleOCR(OcrWrapper):
             Google Cloud Vision API.
     """
 
-    @requires_gcloud
     def __init__(
         self,
         *,
@@ -197,6 +177,11 @@ def __init__(
         add_qr_barcodes: bool = False,
         verbose: bool = False,
     ):
+        try:
+            from google.cloud import vision
+        except ImportError:
+            raise ImportError('GoogleOCR requires missing "google-cloud-vision" package.')
+
         super().__init__(
             cache_file=cache_file,
             max_size=max_size,
@@ -219,11 +204,15 @@ def __init__(
         self.endpoint = endpoint
         self.client = vision.ImageAnnotatorClient(client_options={"api_endpoint": self.endpoint})
 
-    @requires_gcloud
     @tracer.start_as_current_span(name="GoogleOCR._get_ocr")
     def _get_ocr_response(self, img: Image.Image):
         """Gets the OCR response from the Google cloud. Uses cached response if a cache file has been specified and the
         document has been OCRed already"""
+        try:
+            from google.cloud import vision
+        except ImportError:
+            raise ImportError('GoogleOCR requires missing "google-cloud-vision" package.')
+
         span = trace.get_current_span()
         set_image_attributes(span, img)
 
@@ -257,7 +246,6 @@ def _get_ocr_response(self, img: Image.Image):
             self._put_on_shelf(img, response)
         return response
 
-    @requires_gcloud
     @tracer.start_as_current_span(name="GoogleOCR._convert_ocr_response")
     def _convert_ocr_response(self, response) -> tuple[List[BBox], dict[str, Any]]:
         """Converts the response given by Google OCR to a list of BBox"""