deepdoctection 0.35__py3-none-any.whl → 0.37__py3-none-any.whl

deepdoctection/__init__.py

@@ -24,14 +24,10 @@ from .utils.logger import LoggingRecord, logger
 
 # pylint: enable=wrong-import-position
 
-__version__ = 0.35
+__version__ = 0.37
 
 _IMPORT_STRUCTURE = {
-    "analyzer": [
-        "config_sanity_checks",
-        "get_dd_analyzer",
-        "ServiceFactory"
-    ],
+    "analyzer": ["config_sanity_checks", "get_dd_analyzer", "ServiceFactory"],
     "configs": [],
     "dataflow": [
         "DataFlowTerminated",
@@ -91,6 +87,7 @@ _IMPORT_STRUCTURE = {
         "convert_b64_to_np_array",
         "convert_np_array_to_b64",
         "convert_np_array_to_b64_b",
+        "convert_bytes_to_np_array",
         "convert_pdf_bytes_to_np_array_v2",
         "box_to_point4",
         "point4_to_box",
@@ -375,10 +372,12 @@ _IMPORT_STRUCTURE = {
         "save_config_to_yaml",
         "config_to_cli_str",
         "decrypt_pdf_document",
+        "decrypt_pdf_document_from_bytes",
         "get_pdf_file_reader",
         "get_pdf_file_writer",
         "PDFStreamer",
         "pdf_to_np_array",
+        "split_pdf",
         "ObjectTypes",
         "TypeOrStr",
         "object_types_registry",

deepdoctection/analyzer/_config.py

@@ -17,11 +17,13 @@
 
 """Pipeline configuration for deepdoctection analyzer. Do not change the defaults in this file. """
 
+from ..datapoint.view import IMAGE_DEFAULTS
 from ..utils.metacfg import AttrDict
 from ..utils.settings import CellType, LayoutType
 
 cfg = AttrDict()
 
+
 cfg.LANGUAGE = None
 cfg.LIB = None
 cfg.DEVICE = None
@@ -32,11 +34,9 @@ cfg.USE_TABLE_SEGMENTATION = True
 cfg.TF.LAYOUT.WEIGHTS = "layout/model-800000_inf_only.data-00000-of-00001"
 cfg.TF.LAYOUT.FILTER = None
 
-
 cfg.TF.CELL.WEIGHTS = "cell/model-1800000_inf_only.data-00000-of-00001"
 cfg.TF.CELL.FILTER = None
 
-
 cfg.TF.ITEM.WEIGHTS = "item/model-1620000_inf_only.data-00000-of-00001"
 cfg.TF.ITEM.FILTER = None
 
@@ -112,7 +112,7 @@ cfg.OCR.WEIGHTS.DOCTR_WORD.PT = "doctr/db_resnet50/pt/db_resnet50-ac60cadc.pt"
 cfg.OCR.WEIGHTS.DOCTR_RECOGNITION.TF = "doctr/crnn_vgg16_bn/tf/crnn_vgg16_bn-76b7f2c6.zip"
 cfg.OCR.WEIGHTS.DOCTR_RECOGNITION.PT = "doctr/crnn_vgg16_bn/pt/crnn_vgg16_bn-9762b0b0.pt"
 
-cfg.TEXT_CONTAINER = LayoutType.WORD
+cfg.TEXT_CONTAINER = IMAGE_DEFAULTS["text_container"]
 cfg.WORD_MATCHING.PARENTAL_CATEGORIES = [
     LayoutType.TEXT,
     LayoutType.TITLE,
@@ -127,24 +127,16 @@ cfg.WORD_MATCHING.RULE = "ioa"
 cfg.WORD_MATCHING.THRESHOLD = 0.6
 cfg.WORD_MATCHING.MAX_PARENT_ONLY = True
 
-cfg.TEXT_ORDERING.TEXT_BLOCK_CATEGORIES = [
-    LayoutType.TEXT,
-    LayoutType.TITLE,
-    LayoutType.LIST,
-    LayoutType.CELL,
-    CellType.COLUMN_HEADER,
-    CellType.PROJECTED_ROW_HEADER,
-    CellType.SPANNING,
-    CellType.ROW_HEADER,
-]
-cfg.TEXT_ORDERING.FLOATING_TEXT_BLOCK_CATEGORIES = [
-    LayoutType.TEXT,
-    LayoutType.TITLE,
-    LayoutType.LIST,
-]
+cfg.TEXT_ORDERING.TEXT_BLOCK_CATEGORIES = IMAGE_DEFAULTS["text_block_categories"]
+cfg.TEXT_ORDERING.FLOATING_TEXT_BLOCK_CATEGORIES = IMAGE_DEFAULTS["floating_text_block_categories"]
 cfg.TEXT_ORDERING.INCLUDE_RESIDUAL_TEXT_CONTAINER = False
 cfg.TEXT_ORDERING.STARTING_POINT_TOLERANCE = 0.005
 cfg.TEXT_ORDERING.BROKEN_LINE_TOLERANCE = 0.003
 cfg.TEXT_ORDERING.HEIGHT_TOLERANCE = 2.0
 cfg.TEXT_ORDERING.PARAGRAPH_BREAK = 0.035
+
+cfg.USE_LAYOUT_LINK = False
+cfg.LAYOUT_LINK.PARENTAL_CATEGORIES = []
+cfg.LAYOUT_LINK.CHILD_CATEGORIES = []
+
 cfg.freeze()

deepdoctection/analyzer/factory.py

@@ -33,7 +33,13 @@ from ..extern.tessocr import TesseractOcrDetector, TesseractRotationTransformer
 from ..extern.texocr import TextractOcrDetector
 from ..extern.tpdetect import TPFrcnnDetector
 from ..pipe.base import PipelineComponent
-from ..pipe.common import AnnotationNmsService, IntersectionMatcher, MatchingService, PageParsingService
+from ..pipe.common import (
+    AnnotationNmsService,
+    IntersectionMatcher,
+    MatchingService,
+    NeighbourMatcher,
+    PageParsingService,
+)
 from ..pipe.doctectionpipe import DoctectionPipe
 from ..pipe.layout import ImageLayoutService
 from ..pipe.order import TextOrderService
@@ -78,7 +84,7 @@ class ServiceFactory:
     """
 
     @staticmethod
-    def build_layout_detector(
+    def _build_layout_detector(
         config: AttrDict,
         mode: str,
     ) -> Union[D2FrcnnDetector, TPFrcnnDetector, HFDetrDerivedDetector, D2FrcnnTracingDetector]:
@@ -141,17 +147,38 @@ class ServiceFactory:
         )
 
     @staticmethod
-    def build_rotation_detector() -> TesseractRotationTransformer:
+    def build_layout_detector(
+        config: AttrDict, mode: str
+    ) -> Union[D2FrcnnDetector, TPFrcnnDetector, HFDetrDerivedDetector, D2FrcnnTracingDetector]:
+        """Building a layout detector according to the config
+
+        :param config: configuration object
+        :param mode: either `LAYOUT`,`CELL` or `ITEM`
+        """
+        return ServiceFactory._build_layout_detector(config, mode)
+
+    @staticmethod
+    def _build_rotation_detector() -> TesseractRotationTransformer:
         """Building a rotation detector"""
         return TesseractRotationTransformer()
 
     @staticmethod
-    def build_transform_service(transform_predictor: ImageTransformer) -> SimpleTransformService:
+    def build_rotation_detector() -> TesseractRotationTransformer:
+        """Building a rotation detector"""
+        return ServiceFactory._build_rotation_detector()
+
+    @staticmethod
+    def _build_transform_service(transform_predictor: ImageTransformer) -> SimpleTransformService:
         """Building a transform service with a given predictor"""
         return SimpleTransformService(transform_predictor)
 
     @staticmethod
-    def build_padder(config: AttrDict, mode: str) -> PadTransform:
+    def build_transform_service(transform_predictor: ImageTransformer) -> SimpleTransformService:
+        """Building a transform service with a given predictor"""
+        return ServiceFactory._build_transform_service(transform_predictor)
+
+    @staticmethod
+    def _build_padder(config: AttrDict, mode: str) -> PadTransform:
         """Building a padder according to the config
 
         :param config: configuration object
@@ -164,10 +191,20 @@ class ServiceFactory:
             getattr(config.PT, mode).PAD.BOTTOM,
             getattr(config.PT, mode).PAD.LEFT,
         )
-        return PadTransform(top=top, right=right, bottom=bottom, left=left)
+        return PadTransform(top=top, right=right, bottom=bottom, left=left)  #
 
     @staticmethod
-    def build_layout_service(config: AttrDict, detector: ObjectDetector, mode: str) -> ImageLayoutService:
+    def build_padder(config: AttrDict, mode: str) -> PadTransform:
+        """Building a padder according to the config
+
+        :param config: configuration object
+        :param mode: either `LAYOUT`,`CELL` or `ITEM`
+        :return `PadTransform` instance
+        """
+        return ServiceFactory._build_padder(config, mode)
+
+    @staticmethod
+    def _build_layout_service(config: AttrDict, detector: ObjectDetector, mode: str) -> ImageLayoutService:
         """Building a layout service with a given detector
 
         :param config: configuration object
@@ -181,7 +218,18 @@ class ServiceFactory:
         return ImageLayoutService(layout_detector=detector, to_image=True, crop_image=True, padder=padder)
 
     @staticmethod
-    def build_layout_nms_service(config: AttrDict) -> AnnotationNmsService:
+    def build_layout_service(config: AttrDict, detector: ObjectDetector, mode: str) -> ImageLayoutService:
+        """Building a layout service with a given detector
+
+        :param config: configuration object
+        :param detector: will be passed to the `ImageLayoutService`
+        :param mode: either `LAYOUT`,`CELL` or `ITEM`
+        :return `ImageLayoutService` instance
+        """
+        return ServiceFactory._build_layout_service(config, detector, mode)
+
+    @staticmethod
+    def _build_layout_nms_service(config: AttrDict) -> AnnotationNmsService:
         """Building a NMS service for layout annotations
 
         :param config: configuration object
@@ -199,7 +247,15 @@ class ServiceFactory:
         )
 
     @staticmethod
-    def build_sub_image_service(config: AttrDict, detector: ObjectDetector, mode: str) -> SubImageLayoutService:
+    def build_layout_nms_service(config: AttrDict) -> AnnotationNmsService:
+        """Building a NMS service for layout annotations
+
+        :param config: configuration object
+        """
+        return ServiceFactory._build_layout_nms_service(config)
+
+    @staticmethod
+    def _build_sub_image_service(config: AttrDict, detector: ObjectDetector, mode: str) -> SubImageLayoutService:
         """
         Building a sub image layout service with a given detector
 
@@ -226,7 +282,19 @@ class ServiceFactory:
         )
 
     @staticmethod
-    def build_ocr_detector(config: AttrDict) -> Union[TesseractOcrDetector, DoctrTextRecognizer, TextractOcrDetector]:
+    def build_sub_image_service(config: AttrDict, detector: ObjectDetector, mode: str) -> SubImageLayoutService:
+        """
+        Building a sub image layout service with a given detector
+
+        :param config: configuration object
+        :param detector: will be passed to the `SubImageLayoutService`
+        :param mode: either `LAYOUT`,`CELL` or `ITEM`
+        :return: `SubImageLayoutService` instance
+        """
+        return ServiceFactory._build_sub_image_service(config, detector, mode)
+
+    @staticmethod
+    def _build_ocr_detector(config: AttrDict) -> Union[TesseractOcrDetector, DoctrTextRecognizer, TextractOcrDetector]:
         """
         Building OCR predictor
 
@@ -266,6 +334,15 @@ class ServiceFactory:
             return TextractOcrDetector(**credentials_kwargs)
         raise ValueError("You have set USE_OCR=True but any of USE_TESSERACT, USE_DOCTR, USE_TEXTRACT is set to False")
 
+    @staticmethod
+    def build_ocr_detector(config: AttrDict) -> Union[TesseractOcrDetector, DoctrTextRecognizer, TextractOcrDetector]:
+        """
+        Building OCR predictor
+
+        :param config: configuration object
+        """
+        return ServiceFactory._build_ocr_detector(config)
+
     @staticmethod
     def build_doctr_word_detector(config: AttrDict) -> DoctrTextlineDetector:
         """Building `DoctrTextlineDetector` instance
@@ -285,7 +362,7 @@ class ServiceFactory:
         )
 
     @staticmethod
-    def build_table_segmentation_service(
+    def _build_table_segmentation_service(
         config: AttrDict,
         detector: ObjectDetector,
     ) -> Union[PubtablesSegmentationService, TableSegmentationService]:
@@ -342,7 +419,32 @@ class ServiceFactory:
         return table_segmentation
 
     @staticmethod
-    def build_table_refinement_service(config: AttrDict) -> TableSegmentationRefinementService:
+    def build_table_segmentation_service(
+        config: AttrDict,
+        detector: ObjectDetector,
+    ) -> Union[PubtablesSegmentationService, TableSegmentationService]:
+        """
+        Build and return a table segmentation service based on the provided detector.
+
+        Depending on the type of the detector, this method will return either a `PubtablesSegmentationService` or a
+        `TableSegmentationService` instance. The selection is made as follows:
+
+        - If the detector is an instance of `HFDetrDerivedDetector`, a `PubtablesSegmentationService` is created and
+          returned. This service uses specific configuration parameters for segmentation, such as assignment rules,
+          thresholds, and cell names defined in the `cfg` object.
+        - For other detector types, a `TableSegmentationService` is created and returned. This service also uses
+          configuration parameters from the `cfg` object but is tailored for different segmentation needs.
+
+        :param config: configuration object
+        :param detector: An instance of `ObjectDetector` used to determine the type of table segmentation
+        service to build.
+        :return: An instance of either `PubtablesSegmentationService` or `TableSegmentationService` based on the
+                 detector type.
+        """
+        return ServiceFactory._build_table_segmentation_service(config, detector)
+
+    @staticmethod
+    def _build_table_refinement_service(config: AttrDict) -> TableSegmentationRefinementService:
         """Building a table segmentation refinement service
 
         :param config: configuration object
@@ -354,7 +456,16 @@ class ServiceFactory:
         )
 
     @staticmethod
-    def build_pdf_text_detector(config: AttrDict) -> PdfPlumberTextDetector:
+    def build_table_refinement_service(config: AttrDict) -> TableSegmentationRefinementService:
+        """Building a table segmentation refinement service
+
+        :param config: configuration object
+        :return: TableSegmentationRefinementService
+        """
+        return ServiceFactory._build_table_refinement_service(config)
+
+    @staticmethod
+    def _build_pdf_text_detector(config: AttrDict) -> PdfPlumberTextDetector:
         """Building a PDF text detector
 
         :param config: configuration object
@@ -365,7 +476,16 @@ class ServiceFactory:
         )
 
     @staticmethod
-    def build_pdf_miner_text_service(detector: PdfMiner) -> TextExtractionService:
+    def build_pdf_text_detector(config: AttrDict) -> PdfPlumberTextDetector:
+        """Building a PDF text detector
+
+        :param config: configuration object
+        :return: PdfPlumberTextDetector
+        """
+        return ServiceFactory._build_pdf_text_detector(config)
+
+    @staticmethod
+    def _build_pdf_miner_text_service(detector: PdfMiner) -> TextExtractionService:
         """Building a PDFMiner text extraction service
 
         :param detector: PdfMiner
@@ -373,6 +493,15 @@ class ServiceFactory:
         """
         return TextExtractionService(detector)
 
+    @staticmethod
+    def build_pdf_miner_text_service(detector: PdfMiner) -> TextExtractionService:
+        """Building a PDFMiner text extraction service
+
+        :param detector: PdfMiner
+        :return: TextExtractionService
+        """
+        return ServiceFactory._build_pdf_miner_text_service(detector)
+
     @staticmethod
     def build_doctr_word_detector_service(detector: DoctrTextlineDetector) -> ImageLayoutService:
         """Building a Doctr word detector service
@@ -385,7 +514,7 @@ class ServiceFactory:
         )
 
     @staticmethod
-    def build_text_extraction_service(
+    def _build_text_extraction_service(
         config: AttrDict, detector: Union[TesseractOcrDetector, DoctrTextRecognizer, TextractOcrDetector]
     ) -> TextExtractionService:
         """Building a text extraction service
@@ -401,7 +530,19 @@ class ServiceFactory:
         )
 
     @staticmethod
-    def build_word_matching_service(config: AttrDict) -> MatchingService:
+    def build_text_extraction_service(
+        config: AttrDict, detector: Union[TesseractOcrDetector, DoctrTextRecognizer, TextractOcrDetector]
+    ) -> TextExtractionService:
+        """Building a text extraction service
+
+        :param config: configuration object
+        :param detector: OCR detector
+        :return: TextExtractionService
+        """
+        return ServiceFactory._build_text_extraction_service(config, detector)
+
+    @staticmethod
+    def _build_word_matching_service(config: AttrDict) -> MatchingService:
         """Building a word matching service
 
         :param config: configuration object
@@ -420,7 +561,40 @@ class ServiceFactory:
         )
 
     @staticmethod
-    def build_text_order_service(config: AttrDict) -> TextOrderService:
+    def build_word_matching_service(config: AttrDict) -> MatchingService:
+        """Building a word matching service
+
+        :param config: configuration object
+        :return: MatchingService
+        """
+        return ServiceFactory._build_word_matching_service(config)
+
+    @staticmethod
+    def _build_layout_link_matching_service(config: AttrDict) -> MatchingService:
+        """Building a word matching service
+
+        :param config: configuration object
+        :return: MatchingService
+        """
+        neighbor_matcher = NeighbourMatcher()
+        return MatchingService(
+            parent_categories=config.LAYOUT_LINK.PARENTAL_CATEGORIES,
+            child_categories=config.LAYOUT_LINK.CHILD_CATEGORIES,
+            matcher=neighbor_matcher,
+            relationship_key=Relationships.LAYOUT_LINK,
+        )
+
+    @staticmethod
+    def build_layout_link_matching_service(config: AttrDict) -> MatchingService:
+        """Building a word matching service
+
+        :param config: configuration object
+        :return: MatchingService
+        """
+        return ServiceFactory._build_layout_link_matching_service(config)
+
+    @staticmethod
+    def _build_text_order_service(config: AttrDict) -> TextOrderService:
         """Building a text order service
 
         :param config: configuration object
@@ -438,7 +612,16 @@ class ServiceFactory:
         )
 
     @staticmethod
-    def build_page_parsing_service(config: AttrDict) -> PageParsingService:
+    def build_text_order_service(config: AttrDict) -> TextOrderService:
+        """Building a text order service
+
+        :param config: configuration object
+        :return: TextOrderService instance
+        """
+        return ServiceFactory._build_text_order_service(config)
+
+    @staticmethod
+    def _build_page_parsing_service(config: AttrDict) -> PageParsingService:
         """Building a page parsing service
 
         :param config: configuration object
@@ -450,6 +633,15 @@ class ServiceFactory:
             include_residual_text_container=config.TEXT_ORDERING.INCLUDE_RESIDUAL_TEXT_CONTAINER,
         )
 
+    @staticmethod
+    def build_page_parsing_service(config: AttrDict) -> PageParsingService:
+        """Building a page parsing service
+
+        :param config: configuration object
+        :return: PageParsingService instance
+        """
+        return ServiceFactory._build_page_parsing_service(config)
+
     @staticmethod
     def build_analyzer(config: AttrDict) -> DoctectionPipe:
         """
@@ -517,6 +709,10 @@ class ServiceFactory:
             text_order_service = ServiceFactory.build_text_order_service(config)
             pipe_component_list.append(text_order_service)
 
+        if config.USE_LAYOUT_LINK:
+            layout_link_matching_service = ServiceFactory.build_layout_link_matching_service(config)
+            pipe_component_list.append(layout_link_matching_service)
+
         page_parsing_service = ServiceFactory.build_page_parsing_service(config)
 
         return DoctectionPipe(pipeline_component_list=pipe_component_list, page_parsing_service=page_parsing_service)

deepdoctection/configs/conf_dd_one.yaml

@@ -98,3 +98,7 @@ TEXT_ORDERING:
   BROKEN_LINE_TOLERANCE: 0.003
   HEIGHT_TOLERANCE: 2.0
   PARAGRAPH_BREAK: 0.035
+  USE_LAYOUT_LINK: False
+  LAYOUT_LINK:
+    PARENTAL_CATEGORIES:
+    CHILD_CATEGORIES:

deepdoctection/dataflow/custom_serialize.py

@@ -593,7 +593,7 @@ class SerializerPdfDoc:
         file_name = os.path.split(path)[1]
         prefix, suffix = os.path.splitext(file_name)
         df: DataFlow
-        df = CustomDataFromIterable(PDFStreamer(path=path), max_datapoints=max_datapoints)
+        df = CustomDataFromIterable(PDFStreamer(path_or_bytes=path), max_datapoints=max_datapoints)
         df = MapData(
             df,
             lambda dp: {

deepdoctection/datapoint/convert.py

@@ -40,6 +40,7 @@ __all__ = [
     "convert_b64_to_np_array",
     "convert_np_array_to_b64",
     "convert_np_array_to_b64_b",
+    "convert_bytes_to_np_array",
     "convert_pdf_bytes_to_np_array_v2",
     "box_to_point4",
     "point4_to_box",
@@ -107,6 +108,16 @@ def convert_np_array_to_b64_b(np_image: PixelValues) -> bytes:
     return viz_handler.encode(np_image)
 
 
+def convert_bytes_to_np_array(image_bytes: bytes) -> PixelValues:
+    """
+    Converts an image in bytes to a numpy array
+
+    :param image_bytes: An image as bytes.
+    :return: numpy array.
+    """
+    return viz_handler.convert_bytes_to_np(image_bytes)
+
+
 @deprecated("Use convert_pdf_bytes_to_np_array_v2", "2022-02-23")
 def convert_pdf_bytes_to_np_array(pdf_bytes: bytes, dpi: Optional[int] = None) -> PixelValues:
     """

deepdoctection/datapoint/image.py

@@ -587,7 +587,7 @@ class Image:
             )
             ann.image.dump(sub_image)
 
-    def remove_image_from_lower_hierachy(self, pixel_values_only: bool = False) -> None:
+    def remove_image_from_lower_hierarchy(self, pixel_values_only: bool = False) -> None:
         """Will remove all images from image annotations."""
         for ann in self.annotations:
             if pixel_values_only:
@@ -717,7 +717,7 @@ class Image:
         else:
             path_json = fspath(path) + ".json"
         if highest_hierarchy_only:
-            self.remove_image_from_lower_hierachy()
+            self.remove_image_from_lower_hierarchy()
         export_dict = self.as_dict()
         export_dict["location"] = fspath(export_dict["location"])
         if not image_to_json: