deepdoctection 0.44.1__py3-none-any.whl → 0.46__py3-none-any.whl

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."
 
 
@@ -263,7 +292,7 @@ def numpy_v1_available() -> bool:
     Returns:
         True if the installed NumPy version is 1, otherwise False
     """
-    major_version = np.__version__.split('.', maxsplit=1)[0]
+    major_version = np.__version__.split(".", maxsplit=1)[0]
     print(f"major version: {major_version}")
     if major_version in (1, "1"):
         return True

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/settings.py

@@ -108,6 +108,7 @@ class DocumentType(ObjectTypes):
     GOVERNMENT_TENDERS = "government_tenders"
     MANUALS = "manuals"
     PATENTS = "patents"
+    BANK_STATEMENT = "bank_statement"
 
 
 @object_types_registry.register("LayoutType")
@@ -296,6 +297,7 @@ class Languages(ObjectTypes):
     BOSNIAN = "bos"
     NORWEGIAN_NOVOSIBIRSK = "nno"
     URDU = "urd"
+    SWAHILI = "swa"
     NOT_DEFINED = "nn"
 
 

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)
+        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)
+        np_array = cv2.imdecode(np_array, cv2.IMREAD_COLOR).astype(np.float32)  # type: ignore
         return np_array.astype(uint8)
 
     @staticmethod
@@ -543,7 +552,7 @@ class VizPackageHandler:
     def _cv2_convert_bytes_to_np(image_bytes: bytes) -> PixelValues:
         np_array = np.frombuffer(image_bytes, np.uint8)
         np_image = cv2.imdecode(np_array, cv2.IMREAD_COLOR)
-        return np_image
+        return np_image  # type: ignore
 
     @staticmethod
     def _pillow_convert_bytes_to_np(image_bytes: bytes) -> PixelValues:

{deepdoctection-0.44.1.dist-info → deepdoctection-0.46.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: deepdoctection
-Version: 0.44.1
+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,>=1.21
+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,>=1.21; 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"
@@ -58,30 +52,28 @@ Requires-Dist: tensorpack==0.11; extra == "tf"
 Requires-Dist: protobuf==3.20.1; extra == "tf"
 Requires-Dist: tensorflow-addons>=0.17.1; extra == "tf"
 Requires-Dist: tf2onnx>=1.9.2; extra == "tf"
-Requires-Dist: python-doctr==0.9.0; extra == "tf"
+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: fasttext-wheel; 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,>=1.21; 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"
@@ -89,15 +81,16 @@ Requires-Dist: tqdm>=4.64.0; extra == "pt"
 Requires-Dist: timm>=0.9.16; extra == "pt"
 Requires-Dist: transformers>=4.48.0; extra == "pt"
 Requires-Dist: accelerate>=0.29.1; extra == "pt"
-Requires-Dist: python-doctr==0.9.0; extra == "pt"
+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: fasttext-wheel; 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"
@@ -183,7 +176,8 @@ It also provides a framework for training, evaluating and inferencing Document A
   [**LiLT**](https://github.com/jpWang/LiLT) and selected
   [**Bert**](https://huggingface.co/docs/transformers/model_doc/xlm-roberta)-style including features like sliding windows.
 - Text mining for native PDFs with [**pdfplumber**](https://github.com/jsvine/pdfplumber),
-- Language detection with [**fastText**](https://github.com/facebookresearch/fastText),
+- Language detection with `papluca/xlm-roberta-base-language-detection`. [**fastText**](https://github.com/facebookresearch/fastText) is still available but
+  but will be removed in a future version.
 - Deskewing and rotating images with [**jdeskew**](https://github.com/phamquiluan/jdeskew).
 - Fine-tuning and evaluation tools.
 - Lot's of [tutorials](https://github.com/deepdoctection/notebooks)
@@ -294,7 +288,7 @@ alt="text" width="40%">
 
 - Linux or macOS. Windows is not supported but there is a [Dockerfile](./docker/pytorch-cpu-jupyter/Dockerfile) available.
 - Python >= 3.9
-- 2.2 \<= PyTorch **or** 2.11 \<= Tensorflow < 2.16. (For lower Tensorflow versions the code will only run on a GPU).
+- 2.6 \<= PyTorch **or** 2.11 \<= Tensorflow < 2.16. (For lower Tensorflow versions the code will only run on a GPU).
   Tensorflow support will be stopped from Python 3.11 onwards.
 - To fine-tune models, a GPU is recommended.
 
@@ -321,7 +315,7 @@ For a simple setup which is enough to parse documents with the default setting,
 
 ```
 pip install transformers
-pip install python-doctr==0.9.0
+pip install python-doctr==0.10.0 # If you use Python 3.10 or higher you can use the latest version.
 pip install deepdoctection
 ```
 
@@ -329,8 +323,9 @@ pip install deepdoctection
 
 ```
 pip install tensorpack
-pip install python-doctr==0.9.0
 pip install deepdoctection
+pip install "numpy>=1.21,<2.0" --upgrade --force-reinstall  # because TF 2.11 does not support numpy 2.0 
+pip install "python-doctr==0.9.0"
 ```
 
 Both setups are sufficient to run the [**introduction notebook**](https://github.com/deepdoctection/notebooks/blob/main/Get_Started.ipynb).