deepdoctection 0.45.0__py3-none-any.whl → 0.46__py3-none-any.whl

deepdoctection/__init__.py

@@ -25,7 +25,7 @@ from .utils.logger import LoggingRecord, logger
 
 # pylint: enable=wrong-import-position
 
-__version__ = "0.45.0"
+__version__ = "0.46"
 
 _IMPORT_STRUCTURE = {
     "analyzer": ["config_sanity_checks", "get_dd_analyzer", "ServiceFactory", "update_cfg_from_defaults"],
@@ -271,6 +271,7 @@ _IMPORT_STRUCTURE = {
         "MultiThreadPipelineComponent",
         "DoctectionPipe",
         "LanguageDetectionService",
+        "skip_if_category_or_service_extracted",
         "ImageLayoutService",
         "LMTokenClassifierService",
         "LMSequenceClassifierService",
@@ -310,12 +311,14 @@ _IMPORT_STRUCTURE = {
         "get_tensorpack_requirement",
         "pytorch_available",
         "get_pytorch_requirement",
+        "pyzmq_available",
         "lxml_available",
         "get_lxml_requirement",
         "apted_available",
         "get_apted_requirement",
         "distance_available",
         "get_distance_requirement",
+        "networkx_available",
         "numpy_v1_available",
         "get_numpy_v1_requirement",
         "transformers_available",

deepdoctection/analyzer/config.py

@@ -526,6 +526,9 @@ cfg.USE_LM_SEQUENCE_CLASS = False
 # Enables a token classification pipeline component, e.g. a LayoutLM or Bert-like model
 cfg.USE_LM_TOKEN_CLASS = False
 
+# Specifies the selection of the rotation model. There are two models available: A rotation estimator
+# based on Tesseract ('tesseract'), and a rotation estimator based on DocTr ('doctr').
+cfg.ROTATOR.MODEL = "tesseract"
 
 # Relevant when LIB = TF. Specifies the layout detection model.
 # This model should detect multiple or single objects across an entire page.

deepdoctection/analyzer/factory.py

@@ -22,13 +22,13 @@
 from __future__ import annotations
 
 from os import environ
-from typing import TYPE_CHECKING, Union
+from typing import TYPE_CHECKING, Literal, Union
 
 from lazy_imports import try_import
 
 from ..extern.base import ImageTransformer, ObjectDetector, PdfMiner
 from ..extern.d2detect import D2FrcnnDetector, D2FrcnnTracingDetector
-from ..extern.doctrocr import DoctrTextlineDetector, DoctrTextRecognizer
+from ..extern.doctrocr import DocTrRotationTransformer, DoctrTextlineDetector, DoctrTextRecognizer
 from ..extern.hfdetr import HFDetrDerivedDetector
 from ..extern.hflayoutlm import (
     HFLayoutLmSequenceClassifier,
@@ -78,6 +78,7 @@ if TYPE_CHECKING:
     from ..extern.hflayoutlm import LayoutSequenceModels, LayoutTokenModels
     from ..extern.hflm import LmSequenceModels, LmTokenModels
 
+    RotationTransformer = Union[TesseractRotationTransformer, DocTrRotationTransformer]
 
 __all__ = [
     "ServiceFactory",
@@ -190,24 +191,32 @@ class ServiceFactory:
         return ServiceFactory._build_layout_detector(config, mode)
 
     @staticmethod
-    def _build_rotation_detector() -> TesseractRotationTransformer:
+    def _build_rotation_detector(rotator_name: Literal["tesseract", "doctr"]) -> RotationTransformer:
         """
         Building a rotation detector.
 
         Returns:
             TesseractRotationTransformer: Rotation detector instance.
         """
-        return TesseractRotationTransformer()
+
+        if rotator_name == "tesseract":
+            return TesseractRotationTransformer()
+        if rotator_name == "doctr":
+            return DocTrRotationTransformer()
+        raise ValueError(
+            f"You have chosen rotator_name: {rotator_name} which is not allowed. Only tesseract or "
+            f"doctr are allowed."
+        )
 
     @staticmethod
-    def build_rotation_detector() -> TesseractRotationTransformer:
+    def build_rotation_detector(rotator_name: Literal["tesseract", "doctr"]) -> RotationTransformer:
         """
         Building a rotation detector.
 
         Returns:
             TesseractRotationTransformer: Rotation detector instance.
         """
-        return ServiceFactory._build_rotation_detector()
+        return ServiceFactory._build_rotation_detector(rotator_name)
 
     @staticmethod
     def _build_transform_service(transform_predictor: ImageTransformer) -> SimpleTransformService:
@@ -1123,7 +1132,7 @@ class ServiceFactory:
         pipe_component_list: list[PipelineComponent] = []
 
         if config.USE_ROTATOR:
-            rotation_detector = ServiceFactory.build_rotation_detector()
+            rotation_detector = ServiceFactory.build_rotation_detector(config.ROTATOR.MODEL)
             transform_service = ServiceFactory.build_transform_service(transform_predictor=rotation_detector)
             pipe_component_list.append(transform_service)
 

deepdoctection/dataflow/parallel_map.py

@@ -24,15 +24,19 @@ from abc import ABC, abstractmethod
 from contextlib import contextmanager
 from typing import Any, Callable, Iterator, no_type_check
 
-import zmq
+from lazy_imports import try_import
 
 from ..utils.concurrency import StoppableThread, enable_death_signal, start_proc_mask_signal
 from ..utils.error import DataFlowTerminatedError
+from ..utils.file_utils import pyzmq_available
 from ..utils.logger import LoggingRecord, logger
 from .base import DataFlow, DataFlowReentrantGuard, ProxyDataFlow
 from .common import RepeatedData
 from .serialize import PickleSerializer
 
+with try_import() as import_guard:
+    import zmq
+
 
 @no_type_check
 def del_weakref(x):
@@ -77,6 +81,8 @@ def _get_pipe_name(name):
 
 class _ParallelMapData(ProxyDataFlow, ABC):
     def __init__(self, df: DataFlow, buffer_size: int, strict: bool = False) -> None:
+        if not pyzmq_available():
+            raise ModuleNotFoundError("pyzmq is required for running parallel dataflows (multiprocess/multithread).")
         super().__init__(df)
         if buffer_size <= 0:
             raise ValueError(f"buffer_size must be a positive number, got {buffer_size}")

deepdoctection/datapoint/box.py

@@ -284,7 +284,7 @@ class BoundingBox:
             raise BoundingBoxError(
                 f"bounding box must have height and width >0. Check coords "
                 f"ulx: {self.ulx}, uly: {self.uly}, lrx: {self.lrx}, "
-                f"lry: {self.lry}."
+                f"lry: {self.lry}, absolute_coords: {self.absolute_coords}"
             )
         if not self.absolute_coords and not (
             0 <= self.ulx <= 1 and 0 <= self.uly <= 1 and 0 <= self.lrx <= 1 and 0 <= self.lry <= 1
@@ -505,10 +505,10 @@ class BoundingBox:
             if self.absolute_coords:
                 transformed_box = BoundingBox(
                     absolute_coords=not self.absolute_coords,
-                    ulx=max(self.ulx / image_width, 0.0),
-                    uly=max(self.uly / image_height, 0.0),
-                    lrx=min(self.lrx / image_width, 1.0),
-                    lry=min(self.lry / image_height, 1.0),
+                    ulx=min(max(self.ulx / image_width, 0.0), 1.0),
+                    uly=min(max(self.uly / image_height, 0.0), 1.0),
+                    lrx=max(min(self.lrx / image_width, 1.0), 0.0),
+                    lry=max(min(self.lry / image_height, 1.0), 0.0),
                 )
             else:
                 transformed_box = BoundingBox(

deepdoctection/datapoint/image.py

@@ -36,7 +36,7 @@ from ..utils.logger import LoggingRecord, logger
 from ..utils.settings import ObjectTypes, SummaryType, get_type
 from ..utils.types import ImageDict, PathLikeOrStr, PixelValues
 from .annotation import Annotation, AnnotationMap, BoundingBox, CategoryAnnotation, ImageAnnotation
-from .box import crop_box_from_image, global_to_local_coords, intersection_box
+from .box import BoxCoordinate, crop_box_from_image, global_to_local_coords, intersection_box
 from .convert import as_dict, convert_b64_to_np_array, convert_np_array_to_b64, convert_pdf_bytes_to_np_array_v2
 
 
@@ -318,7 +318,7 @@ class Image:
         return _Img(self.image)
 
     @property
-    def width(self) -> float:
+    def width(self) -> BoxCoordinate:
         """
         `width`
         """
@@ -327,7 +327,7 @@ class Image:
         return self._bbox.width
 
     @property
-    def height(self) -> float:
+    def height(self) -> BoxCoordinate:
         """
         `height`
         """
@@ -335,7 +335,7 @@ class Image:
             raise ImageError("Height not available. Call set_width_height first")
         return self._bbox.height
 
-    def set_width_height(self, width: float, height: float) -> None:
+    def set_width_height(self, width: BoxCoordinate, height: BoxCoordinate) -> None:
         """
         Defines bounding box of the image if not already set. Use this, if you do not want to keep the image separated
         for memory reasons.
@@ -345,7 +345,7 @@ class Image:
             height: height of image
         """
         if self._bbox is None:
-            self._bbox = BoundingBox(ulx=0.0, uly=0.0, height=height, width=width, absolute_coords=True)
+            self._bbox = BoundingBox(ulx=0, uly=0, height=height, width=width, absolute_coords=True)
             self._self_embedding()
 
     def set_embedding(self, image_id: str, bounding_box: BoundingBox) -> None:

deepdoctection/datapoint/view.py

@@ -428,6 +428,8 @@ class List(Layout):
             A list of words order by reading order. Words with no `reading_order` will not be returned"""
         try:
             list_items = self.list_items
+            if not list_items:
+                return super().get_ordered_words()
             all_words = []
             list_items.sort(key=lambda x: x.bbox[1])
             for list_item in list_items:
@@ -755,6 +757,8 @@ class Table(Layout):
         """
         try:
             cells = self.cells
+            if not cells:
+                return super().get_ordered_words()
             all_words = []
             cells.sort(key=lambda x: (x.ROW_NUMBER, x.COLUMN_NUMBER))
             for cell in cells:
@@ -1054,6 +1058,8 @@ class Page(Image):
         Returns:
             A `Page` instance with all annotations as `ImageAnnotationBaseView` subclasses.
         """
+        if isinstance(image_orig, Page):
+            raise ImageError("Page.from_image() cannot be called on a Page instance.")
 
         if text_container is None:
             text_container = IMAGE_DEFAULTS.TEXT_CONTAINER
@@ -1310,7 +1316,7 @@ class Page(Image):
             If `interactive=False` will return a `np.array`.
         """
 
-        category_names_list: list[Union[str, None]] = []
+        category_names_list: list[Tuple[Union[str, None], Union[str, None]]] = []
         box_stack = []
         cells_found = False
 
@@ -1323,22 +1329,23 @@ class Page(Image):
             anns = self.get_annotation(category_names=list(debug_kwargs.keys()))
             for ann in anns:
                 box_stack.append(self._ann_viz_bbox(ann))
-                category_names_list.append(str(getattr(ann, debug_kwargs[ann.category_name])))
+                val = str(getattr(ann, debug_kwargs[ann.category_name]))
+                category_names_list.append((val, val))
 
         if show_layouts and not debug_kwargs:
             for item in self.layouts:
                 box_stack.append(self._ann_viz_bbox(item))
-                category_names_list.append(item.category_name.value)
+                category_names_list.append((item.category_name.value, item.category_name.value))
 
         if show_figures and not debug_kwargs:
             for item in self.figures:
                 box_stack.append(self._ann_viz_bbox(item))
-                category_names_list.append(item.category_name.value)
+                category_names_list.append((item.category_name.value, item.category_name.value))
 
         if show_tables and not debug_kwargs:
             for table in self.tables:
                 box_stack.append(self._ann_viz_bbox(table))
-                category_names_list.append(LayoutType.TABLE.value)
+                category_names_list.append((LayoutType.TABLE.value, LayoutType.TABLE.value))
                 if show_cells:
                     for cell in table.cells:
                         if cell.category_name in {
@@ -1347,21 +1354,21 @@ class Page(Image):
                         }:
                             cells_found = True
                             box_stack.append(self._ann_viz_bbox(cell))
-                            category_names_list.append(None)
+                            category_names_list.append((None, cell.category_name.value))
                 if show_table_structure:
                     rows = table.rows
                     cols = table.columns
                     for row in rows:
                         box_stack.append(self._ann_viz_bbox(row))
-                        category_names_list.append(None)
+                        category_names_list.append((None, row.category_name.value))
                     for col in cols:
                         box_stack.append(self._ann_viz_bbox(col))
-                        category_names_list.append(None)
+                        category_names_list.append((None, col.category_name.value))
 
         if show_cells and not cells_found and not debug_kwargs:
             for ann in self.get_annotation(category_names=[LayoutType.CELL, CellType.SPANNING]):
                 box_stack.append(self._ann_viz_bbox(ann))
-                category_names_list.append(None)
+                category_names_list.append((None, ann.category_name.value))
 
         if show_words and not debug_kwargs:
             all_words = []
@@ -1379,22 +1386,36 @@ class Page(Image):
                 for word in all_words:
                     box_stack.append(self._ann_viz_bbox(word))
                     if show_token_class:
-                        category_names_list.append(word.token_class.value if word.token_class is not None else None)
+                        category_names_list.append(
+                            (word.token_class.value, word.token_class.value)
+                            if word.token_class is not None
+                            else (None, None)
+                        )
                     else:
-                        category_names_list.append(word.token_tag.value if word.token_tag is not None else None)
+                        category_names_list.append(
+                            (word.token_tag.value, word.token_tag.value) if word.token_tag is not None else (None, None)
+                        )
             else:
                 for word in all_words:
                     if word.token_class is not None and word.token_class != TokenClasses.OTHER:
                         box_stack.append(self._ann_viz_bbox(word))
                         if show_token_class:
-                            category_names_list.append(word.token_class.value if word.token_class is not None else None)
+                            category_names_list.append(
+                                (word.token_class.value, word.token_class.value)
+                                if word.token_class is not None
+                                else (None, None)
+                            )
                         else:
-                            category_names_list.append(word.token_tag.value if word.token_tag is not None else None)
+                            category_names_list.append(
+                                (word.token_tag.value, word.token_tag.value)
+                                if word.token_tag is not None
+                                else (None, None)
+                            )
 
         if show_residual_layouts and not debug_kwargs:
             for item in self.residual_layouts:
                 box_stack.append(item.bbox)
-                category_names_list.append(item.category_name.value)
+                category_names_list.append((item.category_name.value, item.category_name.value))
 
         if self.image is not None:
             scale_fx = scaled_width / self.width

deepdoctection/eval/cocometric.py

@@ -275,6 +275,7 @@ class CocoMetric(MetricBase):
                       get the ultimate F1-score.
             f1_iou: Use with `f1_score=True` and reset the f1 iou threshold
                     per_category: Whether to calculate metrics per category
+            per_category: If set to True, f1 score will be returned by each category.
         """
         if max_detections is not None:
             assert len(max_detections) == 3, max_detections

deepdoctection/extern/base.py

@@ -263,7 +263,7 @@ class PredictorBase(ABC):
         requirements = cls.get_requirements()
         name = cls.__name__ if hasattr(cls, "__name__") else cls.__class__.__name__
         if not all(requirement[1] for requirement in requirements):
-            raise ImportError(
+            raise ModuleNotFoundError(
                 "\n".join(
                     [f"{name} has the following dependencies:"]
                     + [requirement[2] for requirement in requirements if not requirement[1]]
@@ -334,6 +334,11 @@ class DetectionResult:
         block: block number. For reading order from some ocr predictors
         line: line number. For reading order from some ocr predictors
         uuid: uuid. For assigning detection result (e.g. text to image annotations)
+        relationships: A dictionary of relationships. Each key is a relationship type and each value is a list of
+                       uuids of the related annotations.
+        angle: angle of rotation in degrees. Only used for text detection.
+        image_width: image width
+        image_height: image height
     """
 
     box: Optional[list[float]] = None
@@ -348,6 +353,8 @@ class DetectionResult:
     uuid: Optional[str] = None
     relationships: Optional[dict[str, Any]] = None
     angle: Optional[float] = None
+    image_width: Optional[Union[int, float]] = None
+    image_height: Optional[Union[int, float]] = None
 
 
 class ObjectDetector(PredictorBase, ABC):

deepdoctection/extern/doctrocr.py

@@ -24,9 +24,10 @@ from __future__ import annotations
 import os
 from abc import ABC
 from pathlib import Path
-from typing import Any, Literal, Mapping, Optional, Union
+from typing import Any, Literal, Mapping, Optional, Sequence, Union
 from zipfile import ZipFile
 
+import numpy as np
 from lazy_imports import try_import
 
 from ..utils.env_info import ENV_VARS_TRUE
@@ -39,6 +40,7 @@ from ..utils.file_utils import (
 )
 from ..utils.fs import load_json
 from ..utils.settings import LayoutType, ObjectTypes, PageType, TypeOrStr
+from ..utils.transform import RotationTransform
 from ..utils.types import PathLikeOrStr, PixelValues, Requirement
 from ..utils.viz import viz_handler
 from .base import DetectionResult, ImageTransformer, ModelCategories, ObjectDetector, TextRecognizer
@@ -558,12 +560,13 @@ class DocTrRotationTransformer(ImageTransformer):
         """
         Args:
             number_contours: the number of contours used for the orientation estimation
-            ratio_threshold_for_lines: this is the ratio w/h used to discriminates lines
+            ratio_threshold_for_lines: this is the ratio w/h used to discriminate lines
         """
         self.number_contours = number_contours
         self.ratio_threshold_for_lines = ratio_threshold_for_lines
         self.name = "doctr_rotation_transformer"
         self.model_id = self.get_model_id()
+        self.rotator = RotationTransform(360)
 
     def transform_image(self, np_img: PixelValues, specification: DetectionResult) -> PixelValues:
         """
@@ -579,6 +582,19 @@ class DocTrRotationTransformer(ImageTransformer):
         """
         return viz_handler.rotate_image(np_img, specification.angle)  # type: ignore
 
+    def transform_coords(self, detect_results: Sequence[DetectionResult]) -> Sequence[DetectionResult]:
+        if detect_results:
+            if detect_results[0].angle:
+                self.rotator.set_angle(detect_results[0].angle)  # type: ignore
+                self.rotator.set_image_width(detect_results[0].image_width)  # type: ignore
+                self.rotator.set_image_height(detect_results[0].image_height)  # type: ignore
+                transformed_coords = self.rotator.apply_coords(
+                    np.asarray([detect_result.box for detect_result in detect_results], dtype=float)
+                )
+                for idx, detect_result in enumerate(detect_results):
+                    detect_result.box = transformed_coords[idx, :].tolist()
+        return detect_results
+
     def predict(self, np_img: PixelValues) -> DetectionResult:
         angle = estimate_orientation(
             np_img, n_ct=self.number_contours, ratio_threshold_for_lines=self.ratio_threshold_for_lines

deepdoctection/extern/hflayoutlm.py

@@ -1024,12 +1024,9 @@ class HFLayoutLmv2SequenceClassifier(HFLayoutLmSequenceClassifierBase):
         else:
             raise ValueError(f"images must be list but is {type(images)}")
 
-        result = predict_sequence_classes_from_layoutlm(input_ids,
-                                                        attention_mask,
-                                                        token_type_ids,
-                                                        boxes,
-                                                        self.model,
-                                                        images)
+        result = predict_sequence_classes_from_layoutlm(
+            input_ids, attention_mask, token_type_ids, boxes, self.model, images
+        )
 
         result.class_id += 1
         result.class_name = self.categories.categories[result.class_id]
@@ -1123,12 +1120,9 @@ class HFLayoutLmv3SequenceClassifier(HFLayoutLmSequenceClassifierBase):
         else:
             raise ValueError(f"images must be list but is {type(images)}")
 
-        result = predict_sequence_classes_from_layoutlm(input_ids,
-                                                        attention_mask,
-                                                        token_type_ids,
-                                                        boxes,
-                                                        self.model,
-                                                        images)
+        result = predict_sequence_classes_from_layoutlm(
+            input_ids, attention_mask, token_type_ids, boxes, self.model, images
+        )
 
         result.class_id += 1
         result.class_name = self.categories.categories[result.class_id]

deepdoctection/extern/tessocr.py

@@ -28,8 +28,9 @@ from errno import ENOENT
 from itertools import groupby
 from os import environ, fspath
 from pathlib import Path
-from typing import Any, Mapping, Optional, Union
+from typing import Any, Mapping, Optional, Sequence, Union
 
+import numpy as np
 from packaging.version import InvalidVersion, Version, parse
 
 from ..utils.context import save_tmp_file, timeout_manager
@@ -37,6 +38,7 @@ from ..utils.error import DependencyError, TesseractError
 from ..utils.file_utils import _TESS_PATH, get_tesseract_requirement
 from ..utils.metacfg import config_to_cli_str, set_config_by_yaml
 from ..utils.settings import LayoutType, ObjectTypes, PageType
+from ..utils.transform import RotationTransform
 from ..utils.types import PathLikeOrStr, PixelValues, Requirement
 from ..utils.viz import viz_handler
 from .base import DetectionResult, ImageTransformer, ModelCategories, ObjectDetector
@@ -450,6 +452,7 @@ class TesseractRotationTransformer(ImageTransformer):
         self.name = fspath(_TESS_PATH) + "-rotation"
         self.categories = ModelCategories(init_categories={1: PageType.ANGLE})
         self.model_id = self.get_model_id()
+        self.rotator = RotationTransform(360)
 
     def transform_image(self, np_img: PixelValues, specification: DetectionResult) -> PixelValues:
         """
@@ -465,6 +468,19 @@ class TesseractRotationTransformer(ImageTransformer):
         """
         return viz_handler.rotate_image(np_img, specification.angle)  # type: ignore
 
+    def transform_coords(self, detect_results: Sequence[DetectionResult]) -> Sequence[DetectionResult]:
+        if detect_results:
+            if detect_results[0].angle:
+                self.rotator.set_angle(detect_results[0].angle)  # type: ignore
+                self.rotator.set_image_width(detect_results[0].image_width)  # type: ignore
+                self.rotator.set_image_height(detect_results[0].image_height)  # type: ignore
+                transformed_coords = self.rotator.apply_coords(
+                    np.asarray([detect_result.box for detect_result in detect_results], dtype=float)
+                )
+                for idx, detect_result in enumerate(detect_results):
+                    detect_result.box = transformed_coords[idx, :].tolist()
+        return detect_results
+
     def predict(self, np_img: PixelValues) -> DetectionResult:
         """
         Determines the angle of the rotated image. It can only handle angles that are multiples of 90 degrees.

deepdoctection/pipe/order.py

@@ -228,8 +228,8 @@ class OrderGenerator:
         columns: list[BoundingBox] = []
         anns.sort(
             key=lambda x: (
-                x.bounding_box.transform(image_width, image_height).cy,  # type: ignore
-                x.bounding_box.transform(image_width, image_height).cx,  # type: ignore
+                x.get_bounding_box(image_id).transform(image_width, image_height).cy,
+                x.get_bounding_box(image_id).transform(image_width, image_height).cx,
             )
         )
         for ann in anns:
@@ -309,7 +309,9 @@ class OrderGenerator:
         filtered_blocks: Sequence[tuple[int, str]]
         for idx in range(max_block_number + 1):
             filtered_blocks = list(filter(lambda x: x[0] == idx, blocks))  # type: ignore # pylint: disable=W0640
-            sorted_blocks.extend(self._sort_anns_grouped_by_blocks(filtered_blocks, anns, image_width, image_height))
+            sorted_blocks.extend(
+                self._sort_anns_grouped_by_blocks(filtered_blocks, anns, image_width, image_height, image_id)
+            )
         reading_blocks = [(idx + 1, block[1]) for idx, block in enumerate(sorted_blocks)]
 
         if logger.isEnabledFor(DEBUG):
@@ -346,7 +348,11 @@ class OrderGenerator:
 
     @staticmethod
     def _sort_anns_grouped_by_blocks(
-        block: Sequence[tuple[int, str]], anns: Sequence[ImageAnnotation], image_width: float, image_height: float
+        block: Sequence[tuple[int, str]],
+        anns: Sequence[ImageAnnotation],
+        image_width: float,
+        image_height: float,
+        image_id: Optional[str] = None,
     ) -> list[tuple[int, str]]:
         if not block:
             return []
@@ -356,8 +362,8 @@ class OrderGenerator:
         block_anns = [ann for ann in anns if ann.annotation_id in ann_ids]
         block_anns.sort(
             key=lambda x: (
-                round(x.bounding_box.transform(image_width, image_height).uly, 2),  # type: ignore
-                round(x.bounding_box.transform(image_width, image_height).ulx, 2),  # type: ignore
+                round(x.get_bounding_box(image_id).transform(image_width, image_height).uly, 2),
+                round(x.get_bounding_box(image_id).transform(image_width, image_height).ulx, 2),
             )
         )
         return [(block_number, ann.annotation_id) for ann in block_anns]

deepdoctection/pipe/refine.py

@@ -27,7 +27,7 @@ from dataclasses import asdict
 from itertools import chain, product
 from typing import DefaultDict, Optional, Sequence, Union
 
-import networkx as nx  # type: ignore
+from lazy_imports import try_import
 
 from ..datapoint.annotation import ImageAnnotation
 from ..datapoint.box import merge_boxes
@@ -35,10 +35,15 @@ from ..datapoint.image import Image, MetaAnnotation
 from ..extern.base import DetectionResult
 from ..mapper.maputils import MappingContextManager
 from ..utils.error import ImageError
+from ..utils.file_utils import networkx_available
 from ..utils.settings import CellType, LayoutType, ObjectTypes, Relationships, TableType, get_type
 from .base import PipelineComponent
 from .registry import pipeline_component_registry
 
+with try_import() as import_guard:
+    import networkx as nx  # type: ignore
+
+
 __all__ = ["TableSegmentationRefinementService", "generate_html_string"]
 
 
@@ -441,6 +446,10 @@ class TableSegmentationRefinementService(PipelineComponent):
             table_names: Sequence of table object types.
             cell_names: Sequence of cell object types.
         """
+        if not networkx_available():
+            raise ModuleNotFoundError(
+                "TableSegmentationRefinementService requires networkx. Please install separately."
+            )
         self.table_name = table_names
         self.cell_names = cell_names
         super().__init__("table_segment_refine")

deepdoctection/pipe/text.py

@@ -129,6 +129,12 @@ class TextExtractionService(PipelineComponent):
                     width, height = self.predictor.get_width_height(predictor_input)  # type: ignore
 
                 for detect_result in detect_result_list:
+                    if width is not None and height is not None:
+                        box = detect_result.box
+                        if box:
+                            if box[0] >= width or box[1] >= height or box[2] >= width or box[3] >= height:
+                                continue
+
                     if isinstance(self.predictor, TextRecognizer):
                         detect_ann_id = detect_result.uuid
                     else:

deepdoctection/pipe/transform.py

@@ -77,6 +77,9 @@ class SimpleTransformService(PipelineComponent):
                         score=ann.score,
                         class_id=ann.category_id,
                         uuid=ann.annotation_id,
+                        angle=detection_result.angle,
+                        image_width=dp.width,  # we need the original width, not the transformed width
+                        image_height=dp.height,  # same with height
                     )
                 )
             output_detect_results = self.transform_predictor.transform_coords(detect_results)

deepdoctection/utils/file_utils.py

@@ -8,6 +8,7 @@
 Utilities for maintaining dependencies and dealing with external library packages. Parts of this file is adapted from
 <https://github.com/huggingface/transformers/blob/master/src/transformers/file_utils.py>
 """
+import importlib.metadata
 import importlib.util
 import multiprocessing as mp
 import string
@@ -17,7 +18,6 @@ from shutil import which
 from types import ModuleType
 from typing import Any, Union, no_type_check
 
-import importlib_metadata
 import numpy as np
 from packaging import version
 
@@ -72,9 +72,9 @@ def get_tf_version() -> str:
 
         for pkg in candidates:
             try:
-                tf_version = importlib_metadata.version(pkg)
+                tf_version = importlib.metadata.version(pkg)
                 break
-            except importlib_metadata.PackageNotFoundError:
+            except importlib.metadata.PackageNotFoundError:
                 pass
     return tf_version
 
@@ -175,6 +175,19 @@ def get_pytorch_requirement() -> Requirement:
     return "torch", pytorch_available(), _PYTORCH_ERR_MSG
 
 
+_PYZMQ_AVAILABLE = importlib.util.find_spec("zmq") is not None
+
+
+def pyzmq_available() -> bool:
+    """
+    Returns whether pyzmq is installed.
+
+    Returns:
+        bool: True if pyzmq is installed, False otherwise.
+    """
+    return bool(_PYZMQ_AVAILABLE)
+
+
 # lxml
 _LXML_AVAILABLE = importlib.util.find_spec("lxml") is not None
 _LXML_ERR_MSG = f"lxml must be installed. {_GENERIC_ERR_MSG}"
@@ -232,7 +245,7 @@ _DISTANCE_ERR_MSG = f"distance must be installed. {_GENERIC_ERR_MSG}"
 
 def distance_available() -> bool:
     """
-    Returns whether `distance` is available.
+    Returns True if `distance` is available.
 
     Returns:
         bool: `True` if `distance` is available, False otherwise.
@@ -250,6 +263,22 @@ def get_distance_requirement() -> Requirement:
     return "distance", distance_available(), _DISTANCE_ERR_MSG
 
 
+# networkx
+_NETWORKX_AVAILABLE = importlib.util.find_spec("networkx") is not None
+
+
+def networkx_available() -> bool:
+    """
+    Checks if networkx is installed.
+
+    Returns:
+        bool: True if networkx is installed, False otherwise.
+    :return:
+    """
+    return bool(_NETWORKX_AVAILABLE)
+
+
+# numpy
 _NUMPY_V1_ERR_MSG = "numpy v1 must be installed."
 
 

deepdoctection/utils/logger.py

@@ -143,6 +143,43 @@ class FileFormatter(logging.Formatter):
 
 
 _LOG_DIR = None
+
+
+def _coerce_log_level(val: Any) -> Union[int, str]:
+    """Normalize environment log level values.
+
+    Accepts integer values (e.g., ``20``), numeric strings (``"20"``),
+    or names case-insensitively (``"info"``, ``"Warn"``, ...). Returns
+    either an integer level number or a valid uppercase level name
+    accepted by the :mod:`logging` module.
+
+    Args:
+        val: The raw value from the environment variable ``LOG_LEVEL``.
+
+    Returns:
+        int | str: The corresponding logging level as an int or an
+        uppercase string. Defaults to ``"INFO"`` if the input is invalid.
+    """
+    if isinstance(val, int):
+        return val
+    if val is None:
+        return "INFO"
+    s = str(val).strip()
+    if s.isdigit():
+        return int(s)
+    name = s.upper()
+    if name == "WARN":
+        name = "WARNING"
+    if name in logging._nameToLevel:  # pylint: disable=W0212
+        return name
+    lvl = logging.getLevelName(name)
+    return lvl if isinstance(lvl, int) else "INFO"
+
+
+# resolve level from LOG_LEVEL only
+_ENV_LOG_LEVEL = os.environ.get("LOG_LEVEL", "INFO")
+_RESOLVED_LOG_LEVEL = _coerce_log_level(_ENV_LOG_LEVEL)
+
 _CONFIG_DICT: dict[str, Any] = {
     "version": 1,
     "disable_existing_loggers": False,
@@ -155,7 +192,7 @@ _CONFIG_DICT: dict[str, Any] = {
     },
     "root": {
         "handlers": ["streamhandler"],
-        "level": os.environ.get("LOG_LEVEL", "INFO"),
+        "level": _RESOLVED_LOG_LEVEL,
         "propagate": os.environ.get("LOG_PROPAGATE", "False") in ENV_VARS_TRUE,
     },
 }

deepdoctection/utils/transform.py

@@ -408,8 +408,35 @@ class RotationTransform(BaseTransform):
             angle: Angle to rotate the image. Must be one of 90, 180, 270, or 360 degrees.
         """
         self.angle = angle
-        self.image_width: Optional[int] = None
-        self.image_height: Optional[int] = None
+        self.image_width: Optional[Union[int, float]] = None
+        self.image_height: Optional[Union[int, float]] = None
+
+    def set_angle(self, angle: Literal[90, 180, 270, 360]) -> None:
+        """
+        Set angle
+
+        Args:
+            angle: One of 90, 180, 270, or 360 degrees.
+        """
+        self.angle = angle
+
+    def set_image_width(self, image_width: Union[int, float]) -> None:
+        """
+        Set image width
+
+        Args:
+            image_width: Either a positive integer or 1.
+        """
+        self.image_width = image_width
+
+    def set_image_height(self, image_height: Union[int, float]) -> None:
+        """
+        Set image height
+
+        Args:
+            image_height: Either a positive integer or 1.
+        """
+        self.image_height = image_height
 
     def apply_image(self, img: PixelValues) -> PixelValues:
         """
@@ -442,17 +469,16 @@ class RotationTransform(BaseTransform):
             raise ValueError("Initialize image_width and image_height first")
 
         if self.angle == 90:
-            coords[:, [0, 1, 2, 3]] = coords[:, [1, 0, 3, 2]]
+            self.image_width = self.image_height
+            coords[:, [0, 1, 2, 3]] = coords[:, [1, 2, 3, 0]]
             coords[:, [1, 3]] = self.image_width - coords[:, [1, 3]]
-            coords[:, [0, 1, 2, 3]] = coords[:, [0, 3, 2, 1]]
         elif self.angle == 180:
-            coords[:, [0, 2]] = self.image_width - coords[:, [0, 2]]
-            coords[:, [1, 3]] = self.image_height - coords[:, [1, 3]]
-            coords[:, [0, 1, 2, 3]] = coords[:, [2, 3, 0, 1]]
+            coords[:, [0, 2]] = self.image_width - coords[:, [2, 0]]
+            coords[:, [1, 3]] = self.image_height - coords[:, [3, 1]]
         elif self.angle == 270:
-            coords[:, [0, 1, 2, 3]] = coords[:, [1, 0, 3, 2]]
+            self.image_height = self.image_width
+            coords[:, [0, 1, 2, 3]] = coords[:, [3, 0, 1, 2]]
             coords[:, [0, 2]] = self.image_height - coords[:, [0, 2]]
-            coords[:, [0, 1, 2, 3]] = coords[:, [2, 1, 0, 3]]
 
         return coords
 
@@ -473,17 +499,16 @@ class RotationTransform(BaseTransform):
             raise ValueError("Initialize image_width and image_height first")
 
         if self.angle == 90:
-            coords[:, [0, 1, 2, 3]] = coords[:, [1, 0, 3, 2]]
-            coords[:, [0, 2]] = self.image_width - coords[:, [0, 2]]
-            coords[:, [0, 1, 2, 3]] = coords[:, [2, 1, 0, 3]]
+            self.image_height = self.image_width
+            coords[:, [0, 1, 2, 3]] = coords[:, [3, 0, 1, 2]]
+            coords[:, [0, 2]] = self.image_height - coords[:, [0, 2]]
         elif self.angle == 180:
-            coords[:, [0, 2]] = self.image_width - coords[:, [0, 2]]
-            coords[:, [1, 3]] = self.image_height - coords[:, [1, 3]]
-            coords[:, [0, 1, 2, 3]] = coords[:, [2, 3, 0, 1]]
+            coords[:, [0, 2]] = self.image_width - coords[:, [2, 0]]
+            coords[:, [1, 3]] = self.image_height - coords[:, [3, 1]]
         elif self.angle == 270:
-            coords[:, [0, 1, 2, 3]] = coords[:, [1, 0, 3, 2]]
-            coords[:, [1, 3]] = self.image_height - coords[:, [1, 3]]
-            coords[:, [0, 1, 2, 3]] = coords[:, [0, 3, 2, 1]]
+            self.image_width = self.image_height
+            coords[:, [0, 1, 2, 3]] = coords[:, [1, 2, 3, 0]]
+            coords[:, [1, 3]] = self.image_width - coords[:, [1, 3]]
         return coords
 
     def clone(self) -> RotationTransform:

deepdoctection/utils/viz.py

@@ -20,10 +20,11 @@ Visualisation utils. Copied and pasted from
 """
 
 import base64
+import hashlib
 import os
 import sys
 from io import BytesIO
-from typing import Any, Optional, Sequence, no_type_check
+from typing import Any, Optional, Sequence, Tuple, Union, no_type_check
 
 import numpy as np
 import numpy.typing as npt
@@ -177,17 +178,23 @@ _COLORS = (
 )
 
 
-def random_color(rgb: bool = True, maximum: int = 255) -> tuple[int, int, int]:
+def random_color(
+    rgb: bool = True, maximum: int = 255, deterministic_input_str: Optional[str] = None
+) -> tuple[int, int, int]:
     """
     Args:
         rgb: Whether to return RGB colors or BGR colors.
         maximum: Either 255 or 1.
+        deterministic_input_str: A string to use for deterministic color generation.
 
     Returns:
         A tuple of three integers representing the color.
     """
-
-    idx = np.random.randint(0, len(_COLORS))
+    if deterministic_input_str:
+        hash_digest = hashlib.md5(deterministic_input_str.encode("utf-8")).hexdigest()
+        idx = int(hash_digest, 16) % len(_COLORS)
+    else:
+        idx = np.random.randint(0, len(_COLORS))
     ret = _COLORS[idx] * maximum
     if not rgb:
         ret = ret[::-1]
@@ -197,7 +204,7 @@ def random_color(rgb: bool = True, maximum: int = 255) -> tuple[int, int, int]:
 def draw_boxes(
     np_image: PixelValues,
     boxes: npt.NDArray[float32],
-    category_names_list: Optional[list[Optional[str]]] = None,
+    category_names_list: Optional[list[Tuple[Union[str, None], Union[str, None]]]] = None,
     color: Optional[BGR] = None,
     font_scale: float = 1.0,
     rectangle_thickness: int = 4,
@@ -210,7 +217,8 @@ def draw_boxes(
     Args:
         np_image: Image as `np.ndarray`.
         boxes: A numpy array of shape Nx4 where each row is `[x1, y1, x2, y2]`.
-        category_names_list: List of N category names.
+        category_names_list: List of N tuples. The first element is the category name, whereas the second element is
+                             the value, that is going to be displayed in the text box..
         color: A 3-tuple BGR color (in range `[0, 255]`).
         font_scale: Font scale of text box.
         rectangle_thickness: Thickness of bounding box.
@@ -230,13 +238,14 @@ def draw_boxes(
     category_to_color = {}
     if box_color_by_category and category_names_list is not None:
         category_names = set(category_names_list)
-        category_to_color = {category: random_color() for category in category_names}
-
+        category_to_color = {
+            category[1]: random_color(deterministic_input_str=category[1]) for category in category_names
+        }
     boxes = np.array(boxes, dtype="int32")
     if category_names_list is not None:
         assert len(category_names_list) == len(boxes), f"{len(category_names_list)} != {len(boxes)}"
     else:
-        category_names_list = [None] * len(boxes)
+        category_names_list = [(None, None)] * len(boxes)
     areas = (boxes[:, 2] - boxes[:, 0] + 1) * (boxes[:, 3] - boxes[:, 1] + 1)
     sorted_inds = np.argsort(-areas)  # draw large ones first
     assert areas.min() > 0, areas.min()
@@ -255,12 +264,12 @@ def draw_boxes(
         np_image = cv2.cvtColor(np_image, cv2.COLOR_GRAY2BGR).astype(np.uint8)
     for i in sorted_inds:
         box = boxes[i, :]
-        choose_color = category_to_color.get(category_names_list[i]) if category_to_color is not None else color
+        choose_color = category_to_color.get(category_names_list[i][1]) if category_to_color is not None else color
         if choose_color is None:
             choose_color = random_color()
-        if category_names_list[i] is not None:
+        if category_names_list[i][0] is not None:
             np_image = viz_handler.draw_text(
-                np_image, (box[0], box[1]), category_names_list[i], color=choose_color, font_scale=font_scale
+                np_image, (box[0], box[1]), category_names_list[i][0], color=choose_color, font_scale=font_scale
             )
         np_image = viz_handler.draw_rectangle(
             np_image, (box[0], box[1], box[2], box[3]), choose_color, rectangle_thickness
@@ -423,7 +432,7 @@ class VizPackageHandler:
 
     @staticmethod
     def _cv2_read_image(path: PathLikeOrStr) -> PixelValues:
-        return cv2.imread(os.fspath(path), cv2.IMREAD_COLOR).astype(np.uint8) # type: ignore
+        return cv2.imread(os.fspath(path), cv2.IMREAD_COLOR).astype(np.uint8)  # type: ignore
 
     @staticmethod
     def _pillow_read_image(path: PathLikeOrStr) -> PixelValues:
@@ -517,7 +526,7 @@ class VizPackageHandler:
     @staticmethod
     def _cv2_convert_b64_to_np(image: B64Str) -> PixelValues:
         np_array = np.fromstring(base64.b64decode(image), np.uint8)  # type: ignore
-        np_array = cv2.imdecode(np_array, cv2.IMREAD_COLOR).astype(np.float32) # type: ignore
+        np_array = cv2.imdecode(np_array, cv2.IMREAD_COLOR).astype(np.float32)  # type: ignore
         return np_array.astype(uint8)
 
     @staticmethod

{deepdoctection-0.45.0.dist-info → deepdoctection-0.46.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: deepdoctection
-Version: 0.45.0
+Version: 0.46
 Summary: Repository for Document AI
 Home-page: https://github.com/deepdoctection/deepdoctection
 Author: Dr. Janis Meyer
@@ -19,18 +19,15 @@ Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: catalogue==2.0.10
 Requires-Dist: huggingface_hub>=0.26.0
-Requires-Dist: importlib-metadata>=5.0.0
 Requires-Dist: jsonlines==3.1.0
 Requires-Dist: lazy-imports==0.3.1
 Requires-Dist: mock==4.0.3
-Requires-Dist: networkx>=2.7.1
 Requires-Dist: numpy>2.0
 Requires-Dist: packaging>=20.0
 Requires-Dist: Pillow>=10.0.0
 Requires-Dist: pypdf>=6.0.0
 Requires-Dist: pypdfium2>=4.30.0
 Requires-Dist: pyyaml>=6.0.1
-Requires-Dist: pyzmq>=16
 Requires-Dist: scipy>=1.13.1
 Requires-Dist: termcolor>=1.1
 Requires-Dist: tabulate>=0.7.7
@@ -38,18 +35,15 @@ Requires-Dist: tqdm>=4.64.0
 Provides-Extra: tf
 Requires-Dist: catalogue==2.0.10; extra == "tf"
 Requires-Dist: huggingface_hub>=0.26.0; extra == "tf"
-Requires-Dist: importlib-metadata>=5.0.0; extra == "tf"
 Requires-Dist: jsonlines==3.1.0; extra == "tf"
 Requires-Dist: lazy-imports==0.3.1; extra == "tf"
 Requires-Dist: mock==4.0.3; extra == "tf"
-Requires-Dist: networkx>=2.7.1; extra == "tf"
 Requires-Dist: numpy>2.0; extra == "tf"
 Requires-Dist: packaging>=20.0; extra == "tf"
 Requires-Dist: Pillow>=10.0.0; extra == "tf"
 Requires-Dist: pypdf>=6.0.0; extra == "tf"
 Requires-Dist: pypdfium2>=4.30.0; extra == "tf"
 Requires-Dist: pyyaml>=6.0.1; extra == "tf"
-Requires-Dist: pyzmq>=16; extra == "tf"
 Requires-Dist: scipy>=1.13.1; extra == "tf"
 Requires-Dist: termcolor>=1.1; extra == "tf"
 Requires-Dist: tabulate>=0.7.7; extra == "tf"
@@ -62,25 +56,24 @@ Requires-Dist: python-doctr==0.10.0; extra == "tf"
 Requires-Dist: pycocotools>=2.0.2; extra == "tf"
 Requires-Dist: boto3==1.34.102; extra == "tf"
 Requires-Dist: pdfplumber>=0.11.0; extra == "tf"
+Requires-Dist: pyzmq>=16; extra == "tf"
 Requires-Dist: jdeskew>=0.2.2; extra == "tf"
 Requires-Dist: apted==1.0.3; extra == "tf"
 Requires-Dist: distance==0.1.3; extra == "tf"
 Requires-Dist: lxml>=4.9.1; extra == "tf"
+Requires-Dist: networkx>=2.7.1; extra == "tf"
 Provides-Extra: pt
 Requires-Dist: catalogue==2.0.10; extra == "pt"
 Requires-Dist: huggingface_hub>=0.26.0; extra == "pt"
-Requires-Dist: importlib-metadata>=5.0.0; extra == "pt"
 Requires-Dist: jsonlines==3.1.0; extra == "pt"
 Requires-Dist: lazy-imports==0.3.1; extra == "pt"
 Requires-Dist: mock==4.0.3; extra == "pt"
-Requires-Dist: networkx>=2.7.1; extra == "pt"
 Requires-Dist: numpy>2.0; extra == "pt"
 Requires-Dist: packaging>=20.0; extra == "pt"
 Requires-Dist: Pillow>=10.0.0; extra == "pt"
 Requires-Dist: pypdf>=6.0.0; extra == "pt"
 Requires-Dist: pypdfium2>=4.30.0; extra == "pt"
 Requires-Dist: pyyaml>=6.0.1; extra == "pt"
-Requires-Dist: pyzmq>=16; extra == "pt"
 Requires-Dist: scipy>=1.13.1; extra == "pt"
 Requires-Dist: termcolor>=1.1; extra == "pt"
 Requires-Dist: tabulate>=0.7.7; extra == "pt"
@@ -92,10 +85,12 @@ Requires-Dist: python-doctr==0.10.0; extra == "pt"
 Requires-Dist: pycocotools>=2.0.2; extra == "pt"
 Requires-Dist: boto3==1.34.102; extra == "pt"
 Requires-Dist: pdfplumber>=0.11.0; extra == "pt"
+Requires-Dist: pyzmq>=16; extra == "pt"
 Requires-Dist: jdeskew>=0.2.2; extra == "pt"
 Requires-Dist: apted==1.0.3; extra == "pt"
 Requires-Dist: distance==0.1.3; extra == "pt"
 Requires-Dist: lxml>=4.9.1; extra == "pt"
+Requires-Dist: networkx>=2.7.1; extra == "pt"
 Provides-Extra: docs
 Requires-Dist: tensorpack==0.11; extra == "docs"
 Requires-Dist: boto3==1.34.102; extra == "docs"

{deepdoctection-0.45.0.dist-info → deepdoctection-0.46.dist-info}/RECORD

@@ -1,9 +1,9 @@
-deepdoctection/__init__.py,sha256=nyR805N1k7HNYBtc8gqshSBvRxpK0JTKHWchetqQjno,13125
+deepdoctection/__init__.py,sha256=mFZGAnUHzhgUshyF-ggZMExtd4ZyCdtg1-hriRty8Ek,13229
 deepdoctection/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 deepdoctection/analyzer/__init__.py,sha256=wg0BcFwdCeREwzZfa--Yx8HUJ9LPv5z5PmLwtkZdPH8,772
-deepdoctection/analyzer/config.py,sha256=_JqF0-6G-IQ9BmQit9OiUMhZDw7pd-eWlC2RPVpwVWQ,43932
+deepdoctection/analyzer/config.py,sha256=5abZTB_U_FaZmNVkKUexSVZSmx_93Uj28VB-349jrCs,44153
 deepdoctection/analyzer/dd.py,sha256=2BGvZpl9o9khcaOV52-DPHMrs0DsqUO8cpdqFVHHzDQ,5176
-deepdoctection/analyzer/factory.py,sha256=DVrXVN-h0apWQG3shmmXwyvAcqLKIvMsW8bNG5cwI5s,47668
+deepdoctection/analyzer/factory.py,sha256=s4RjPgPhB8_raiUbEyACRuF9S_DRQzAczWKDRmzsyhc,48189
 deepdoctection/configs/__init__.py,sha256=TX_P6tqDOF1LK1mi9ruAl7x0mtv1Asm8cYWCz3Pe2dk,646
 deepdoctection/configs/conf_dd_one.yaml,sha256=DHqAIKH3jRam54QO7qib2zutmpyFA8TqdV5UvIV191A,3688
 deepdoctection/configs/conf_tesseract.yaml,sha256=oF6szDyoi15FHvq7yFUNIEjfA_jNLhGxoowiRsz_zY4,35
@@ -13,15 +13,15 @@ deepdoctection/dataflow/base.py,sha256=ZLRijyHI1J7tBfnE-q7eqUieYMMERjtK-c1oK40dB
 deepdoctection/dataflow/common.py,sha256=DKD_pRZBCt2vO3oNZcOvdoC3jThabTNcNbTS16mpVR0,10351
 deepdoctection/dataflow/custom.py,sha256=xlw1Op4J3a8PNIlzY5stAY6olpBTN8KMhj1KQ7gf8tA,6792
 deepdoctection/dataflow/custom_serialize.py,sha256=zWDx1_mkPpeot9VN-4P1C2sFtK6vYUaFoSs6UiiLMZA,23234
-deepdoctection/dataflow/parallel_map.py,sha256=Xhem9lvNDKVd_x02Ih9qB4J6bEnxNbc8uHIro8mX9UU,15783
+deepdoctection/dataflow/parallel_map.py,sha256=8cTaj9tJjpPFHYtfNNaekQZCywIVc3aa2hMHa5nsFbA,16058
 deepdoctection/dataflow/serialize.py,sha256=G5kfkFHyhy3E3AusTvTizBi0EVDU2YZov9a-LmXPjy4,4592
 deepdoctection/dataflow/stats.py,sha256=AN5cbagveaDGqCXaYj6iWITpn-a2eO_AumA-vQNQ_XE,9764
 deepdoctection/datapoint/__init__.py,sha256=ruyV4DTOkUFWhkJ5VO_eJmrAxrzgygzTtr4W-WkZybs,1615
 deepdoctection/datapoint/annotation.py,sha256=f32BNmzUGJoNMeGst2RGC2jmjJpzzjxyBRKFG8FCubY,23092
-deepdoctection/datapoint/box.py,sha256=QAS8sK2Ge4_ysW6zOYkLlzNwhSyw_mhYcYsxscClEno,31453
+deepdoctection/datapoint/box.py,sha256=rJUcQfCIwIpYsyifxMZ_r-NNRpX-itISATA3OKmVyak,31533
 deepdoctection/datapoint/convert.py,sha256=6ENXX3tBdY8ogb2NBPxsOsQMGnQux8ol5nrUfWS5tYE,7352
-deepdoctection/datapoint/image.py,sha256=N5VH2oeKQWIt5FQvFaeu-FL8eckv7LQS0ZJsHSuVwjI,37187
-deepdoctection/datapoint/view.py,sha256=x7BuWWHWOMQa_dZiKxlNIoq1NxZ-Z2F9nOW6vsJOxwE,61910
+deepdoctection/datapoint/image.py,sha256=kH3dPQlI7-Q1lA40cjnp5FT84z03GSsiwoVmWyhN7Qo,37230
+deepdoctection/datapoint/view.py,sha256=m7R9XmH4qX6IOr1hNSog6h5JDx07DAM2GIEk6rDjSbs,63022
 deepdoctection/datasets/__init__.py,sha256=4ifjIwWCPYiS31GzUlVDScrkNOrb1eo5xHlRXNyg_58,994
 deepdoctection/datasets/adapter.py,sha256=VSLM_980aHi4TpgOxfxiBHiF_fUXyh348PXet6zTo-4,7779
 deepdoctection/datasets/base.py,sha256=oLv2o9QiKVN44kO7Llj-z00_TQRYBsVlvBL3ZQoscUQ,30670
@@ -45,23 +45,23 @@ deepdoctection/datasets/instances/xsl/pascal_voc.xsl,sha256=DlzFV2P8NtQKXVe96i-m
 deepdoctection/eval/__init__.py,sha256=deGj63ejU9f3nthBU6GI25QIQidKWJmIW4q8fpn12bU,920
 deepdoctection/eval/accmetric.py,sha256=TlOFUU9y9BBjJKVsRMyoVKpLZl6AflNsZ4thqSEie4k,19957
 deepdoctection/eval/base.py,sha256=mYVvzD_wVPwsrBqcl1O4Vqqhg1yGtlG6hkuMjVZvt-k,5290
-deepdoctection/eval/cocometric.py,sha256=H-BsLeV9S93tG6jzUN-3FCPXYiUUoTAYuznE5SvS9Bc,11070
+deepdoctection/eval/cocometric.py,sha256=xB7M5w4VFI67EPfsmS92EpTlayQRqjIiHS_T0ENsfq4,11156
 deepdoctection/eval/eval.py,sha256=UUL-wk39RONLMOOyH3WjjpHunZJiQluXZFqir8eaDtY,19808
 deepdoctection/eval/registry.py,sha256=us6EGN_tAia1Mk1mwWQwDeE-xqxcuopztdi8n-ieGbg,1100
 deepdoctection/eval/tedsmetric.py,sha256=EcNeJynsmxyl5bOH3bjy2wE647ONf0SF5OZyGbVu35Q,9963
 deepdoctection/eval/tp_eval_callback.py,sha256=lqrOn2tdaRiF_Vr_9CwBr2ryatcWu3mQKya8YZ2pA9A,5261
 deepdoctection/extern/__init__.py,sha256=1RVkuC0MPlz_g4nhU-nc0sPIRR72JWeDgZtyy4BWw8w,1011
-deepdoctection/extern/base.py,sha256=LomTR9HXcBU55MPDIA8D1rIamk7DUmToJmgcRXzCoeU,31650
+deepdoctection/extern/base.py,sha256=8bhTb4PiZlSbSZYnH65FE55juq6hM1TzbCkZHoIUE8w,32077
 deepdoctection/extern/d2detect.py,sha256=O8XN_sUrQThMmd9-t97lzZvTGVSMG-1DD_VR5TV9V8c,22375
 deepdoctection/extern/deskew.py,sha256=L_jU0rXh03qzwaT79EIqE_zYMUVeFwWDbsGbtahuL2k,3124
-deepdoctection/extern/doctrocr.py,sha256=jB0mnvGmmygoUu9e9zw2_HtAgQUdCJHbxMSt1cfK5bA,25381
+deepdoctection/extern/doctrocr.py,sha256=vF-OA9UzH_NizkSoATvCdGEKbFReThGenoJOKFCHMUA,26299
 deepdoctection/extern/fastlang.py,sha256=0nBFZTwMS5s9fhjgMc_p0y18V6wZwQme0r6B6B3uFro,4952
 deepdoctection/extern/hfdetr.py,sha256=N3eLNI5BsQS9_7YZyBeWndSgUydJij7ugZA9p4V1xaQ,14316
-deepdoctection/extern/hflayoutlm.py,sha256=htPfwwJ5VpzYP6CZ86YHwNfvJ7fdhDF_rxTKUa8rG9w,60488
+deepdoctection/extern/hflayoutlm.py,sha256=ZXXX9BTF5zjcyZaDbhGizfDC_xpvD2LyEivElwYuIGc,59972
 deepdoctection/extern/hflm.py,sha256=ftr5jLb39521KtHZOEyTWuaE7bnbSwm3EQSHlcynXIM,27585
 deepdoctection/extern/model.py,sha256=kMIlx07_kdwZHLYB3QUG0DT_VSv2aZuKIIbv3fs0WqA,18233
 deepdoctection/extern/pdftext.py,sha256=ljzPQn3yYAlS6MoZqzixD-fO2GlHwu1aMiOQ6qMIzbg,7513
-deepdoctection/extern/tessocr.py,sha256=SuPmngsJg38riL4b09z6_FIzJH6H3RIwoighG2GPMYM,17457
+deepdoctection/extern/tessocr.py,sha256=r2UTxOCAHtZP863urD5IwTrJxKk7C9fhv5p5suMTCHA,18376
 deepdoctection/extern/texocr.py,sha256=wVOuu6eUGao0mUbC8vrgdCsKfY1GqA1Am9560YgWyXU,5915
 deepdoctection/extern/tpdetect.py,sha256=Kr00n80V_OfE-EGfpjiVw1eAQ2n2tuT-hSco-dLSR9E,8516
 deepdoctection/extern/pt/__init__.py,sha256=3Cu0ZHjbYsJomru7-RQXEHihEQLegZrmLetlHiqS58I,742
@@ -111,13 +111,13 @@ deepdoctection/pipe/doctectionpipe.py,sha256=ik5F92F3klI5Nve_AnyIRj-ApMoKHSR2Sjc
 deepdoctection/pipe/language.py,sha256=VZvw1hYrs4F1g2aSmqt16jAzgigedui2dptcRtvASfY,5949
 deepdoctection/pipe/layout.py,sha256=pm53RUyMCERHJVWRJmeDUfjf-6DlRuTtUGETHpyr1UY,6391
 deepdoctection/pipe/lm.py,sha256=XtvaqjPK-2exWOmzznCZORL5MjkP-33fd0MlJpTtbMA,20617
-deepdoctection/pipe/order.py,sha256=m31RLoQNTpUTMpuyrAZKcTnRhyPLZ_Bmb1Ngxs7JkbY,41129
-deepdoctection/pipe/refine.py,sha256=AazkdLz5F1H8OIO8d1oFY4pqOprP0zW42ZeXVfeUtew,23422
+deepdoctection/pipe/order.py,sha256=Sfp4SVbQ9LZNjCayZfwV4MObUQTopBK_zIXBCBp69ZQ,41226
+deepdoctection/pipe/refine.py,sha256=KhTiTwQTleUJrDE6MO094jq6p_9POxOyYdovejGnJtM,23738
 deepdoctection/pipe/registry.py,sha256=uT5fnHjffoNGk2JPuD2-pMYtO3Iko7-wrwVZVCWLtok,906
 deepdoctection/pipe/segment.py,sha256=ACMHUDigyEMuNXp8crwFXbWCew1-oAdlAa5fV7PfL50,61964
 deepdoctection/pipe/sub_layout.py,sha256=YH05W24yLvAdEwC5CLGiVSOLlo0O-mNeIjRZPd9FEIQ,14135
-deepdoctection/pipe/text.py,sha256=4iRk5K18WxB2VUSARALrmfpj1hR0CyVoj-NJW7O7hHA,11150
-deepdoctection/pipe/transform.py,sha256=KU5WQ-90AaltSQLuRz0D91OkhSkb6hHRNfSNMW3ANeo,4716
+deepdoctection/pipe/text.py,sha256=l4OoAFWpY7wp_oZplzuA7BOs52BkEXIDWiIsq6zlBK0,11444
+deepdoctection/pipe/transform.py,sha256=LG5qpbTK0sdnIRYU8upqxAaLcA497CHo-rXpRnssQgc,4941
 deepdoctection/train/__init__.py,sha256=YFTRAZF1F7cEAKTdAIi1BLyYb6rSRcwq09Ui5Lu8d6E,1071
 deepdoctection/train/d2_frcnn_train.py,sha256=edmyNTBRMM_TuL_1D6G2TSY9CBqNndIuyKree_KAso0,15508
 deepdoctection/train/hf_detr_train.py,sha256=El-VHggdBObttFQwFIfQs5xm7aaxpC5IzNUJ1gF4Z6E,13278
@@ -129,21 +129,21 @@ deepdoctection/utils/context.py,sha256=5QfdzxsiSPnNs1qtJdgjguIoD8srLQ2W8oeDzwp9F
 deepdoctection/utils/develop.py,sha256=4myrqBDypM6tQ2a2Jo3Q20RuE_W2czykpXBwgXPrxNw,3568
 deepdoctection/utils/env_info.py,sha256=b1WohrfQuoL-BPN0_s8Rjtwzx-WKvCyaX2I4qYl1Emc,19878
 deepdoctection/utils/error.py,sha256=sIry8F5MZ0yLvKfAwVz90IorKWVvjoRqcC0L8qq8mLk,2480
-deepdoctection/utils/file_utils.py,sha256=PzUAE7eaiPl-m4SKXNF5_s3Ks7B0WeolmRaIJ7FNO2U,26276
+deepdoctection/utils/file_utils.py,sha256=Qr6Q9ruzi7xMDYh5W1V2_9HZvn2-VzSjH0eu6dhJTpY,26844
 deepdoctection/utils/fs.py,sha256=KTS9FJzZk9le_vmIPr9IisJw0AyTfjkyX1KoWQy4DNs,12729
 deepdoctection/utils/identifier.py,sha256=Jt12MeZf7eC1qciY5Fp_AYUGxYVcjsy7xNBUvJil7dU,2270
-deepdoctection/utils/logger.py,sha256=ddQ0xBStluf8OvoRlEB8YkqyRR-ZYgyJYLClTmJJMAU,10290
+deepdoctection/utils/logger.py,sha256=Mj7FA4B879rYBFbANaoH0UwP4oxLOPt7zxqobKJ4910,11418
 deepdoctection/utils/metacfg.py,sha256=5M390--ZMoyJEt5oZOwFMGt2i8OF_ayeb0NVmUO_3OQ,7235
 deepdoctection/utils/mocks.py,sha256=IkN3-IzAl4eX0ibgKIHg8IY7ykVw6BnpF6XnxKnKaZI,2389
 deepdoctection/utils/pdf_utils.py,sha256=BrxTuY9j0COyIRkJchJ0tt2h6ZsA2an6z-H8E8QwgUQ,13490
 deepdoctection/utils/settings.py,sha256=nzD2OMxfsL50CvKGnbwn8IWW-t5wGfCS439HFian274,12920
 deepdoctection/utils/tqdm.py,sha256=kx3Ivf0x85S0ZmEaN5mImu0V6isOgygOU8iyr2U99XU,1850
-deepdoctection/utils/transform.py,sha256=jgeCyQWLN9q79jCGW7jysyKUKcJ1AVMk8OslF-3fbag,16095
+deepdoctection/utils/transform.py,sha256=2naYQSyIhVmqhocoYgMZA6mSC-XJb8l_fYMWvJFR1oo,16624
 deepdoctection/utils/types.py,sha256=Nsr2J7XSZazXho94y0oc01LBQxh0ve67c4Yx2gMlSXU,2952
 deepdoctection/utils/utils.py,sha256=NBUb1qbx8Jm-AvYN1Sdbk0huXhbAKxZ-ZtOcMespsMM,7064
-deepdoctection/utils/viz.py,sha256=Aduyr65LoI4l9Fv4HCm4Sz9Fa_rL5mR5mQwLN4rqLdM,27385
-deepdoctection-0.45.0.dist-info/licenses/LICENSE,sha256=GQ0rUvuGdrMNEI3iHK5UQx6dIMU1QwAuyXsxUHn5MEQ,11351
-deepdoctection-0.45.0.dist-info/METADATA,sha256=1obCXEh1LnL5mDI_95SHiu5KrehKqP2r3SL5j4Hm2OE,14972
-deepdoctection-0.45.0.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-deepdoctection-0.45.0.dist-info/top_level.txt,sha256=hs2DdoOL9h4mnHhmO82BT4pz4QATIoOZ20PZmlnxFI8,15
-deepdoctection-0.45.0.dist-info/RECORD,,
+deepdoctection/utils/viz.py,sha256=C_zHbFhn9rpspj9GdS9rffLap2x_cpzqdepd5xbkoAE,27990
+deepdoctection-0.46.dist-info/licenses/LICENSE,sha256=GQ0rUvuGdrMNEI3iHK5UQx6dIMU1QwAuyXsxUHn5MEQ,11351
+deepdoctection-0.46.dist-info/METADATA,sha256=dATdyi6eLfJfY1y0GYVhxY3Yttfvl0wSwjwBKgWvMAw,14761
+deepdoctection-0.46.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+deepdoctection-0.46.dist-info/top_level.txt,sha256=hs2DdoOL9h4mnHhmO82BT4pz4QATIoOZ20PZmlnxFI8,15
+deepdoctection-0.46.dist-info/RECORD,,