deepdoctection 1.0.7__tar.gz → 1.2.0__tar.gz

{deepdoctection-1.0.7 → deepdoctection-1.2.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: deepdoctection
-Version: 1.0.7
+Version: 1.2.0
 Summary: Repository for Document AI - server/inference core package
 Author: Dr. Janis Meyer
 License: Apache License 2.0
@@ -18,7 +18,7 @@ Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
 Requires-Python: >=3.10
 Description-Content-Type: text/markdown
 Requires-Dist: dd-core[full]>=1.0.1
-Requires-Dist: huggingface_hub>=0.26.0
+Requires-Dist: huggingface_hub>=1.0
 Provides-Extra: full
 Requires-Dist: dd-datasets[full]>=1.0.1; extra == "full"
 Requires-Dist: boto3==1.34.102; extra == "full"
@@ -30,7 +30,7 @@ Requires-Dist: distance==0.1.3; extra == "full"
 Requires-Dist: lxml>=4.9.1; extra == "full"
 Requires-Dist: pycocotools>=2.0.2; extra == "full"
 Requires-Dist: timm>=0.9.16; extra == "full"
-Requires-Dist: transformers<5.0.0,>=4.48.0; extra == "full"
+Requires-Dist: transformers>=5.2.0; extra == "full"
 Requires-Dist: accelerate>=0.29.1; extra == "full"
 Requires-Dist: python-doctr>=1.0.0; extra == "full"
 Provides-Extra: types

{deepdoctection-1.0.7 → deepdoctection-1.2.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "deepdoctection"
-version = "1.0.7"
+version = "1.2.0"
 authors = [
     {name = "Dr. Janis Meyer"}
 ]
@@ -25,7 +25,7 @@ classifiers = [
 
 dependencies = [
     "dd-core[full]>=1.0.1",
-    "huggingface_hub>=0.26.0",
+    "huggingface_hub>=1.0",
 ]
 
 [project.optional-dependencies]
@@ -43,7 +43,7 @@ full = [
     "pycocotools>=2.0.2",
     # DL dependencies
     "timm>=0.9.16",
-    "transformers>=4.48.0,<5.0.0",
+    "transformers>=5.2.0",
     "accelerate>=0.29.1",
     "python-doctr>=1.0.0",
 ]

{deepdoctection-1.0.7 → deepdoctection-1.2.0}/src/deepdoctection/__init__.py

@@ -6,13 +6,13 @@ Init file for deepdoctection package. This file is used to import all submodules
 """
 
 import sys
-from typing import TYPE_CHECKING, Dict, List
+from typing import TYPE_CHECKING, Dict
 
 from dd_core.utils.env_info import collect_env_info
 from dd_core.utils.file_utils import _LazyModule
 from dd_core.utils.logger import LoggingRecord, logger
 
-__version__ = "1.0.7"
+__version__ = "1.2.0"
 _IMPORT_STRUCTURE = {
     "analyzer": ["config_sanity_checks", "get_dd_analyzer", "ServiceFactory", "update_cfg_from_defaults"],
     "eval": [

{deepdoctection-1.0.7 → deepdoctection-1.2.0}/src/deepdoctection/analyzer/config.py

@@ -423,7 +423,7 @@ The distance is calculated using the center points of the layout elements.
 
 from dd_core.datapoint.view import IMAGE_DEFAULTS
 from dd_core.utils.metacfg import AttrDict
-from dd_core.utils.object_types import CellType, LayoutType
+from dd_core.utils.object_types import CellKey, CellLabel, LayoutLabel
 
 cfg = AttrDict()
 
@@ -551,33 +551,33 @@ cfg.LAYOUT.PAD.LEFT = 0
 # LAYOUT_NMS_PAIRS.THRESHOLDS = [0.001, 0.01]
 # LAYOUT_NMS_PAIRS.PRIORITY = ['table', None]
 cfg.LAYOUT_NMS_PAIRS.COMBINATIONS = [
-    [LayoutType.TABLE, LayoutType.TITLE],
-    [LayoutType.TABLE, LayoutType.TEXT],
-    [LayoutType.TABLE, LayoutType.KEY_VALUE_AREA],
-    [LayoutType.TABLE, LayoutType.LIST_ITEM],
-    [LayoutType.TABLE, LayoutType.LIST],
-    [LayoutType.TABLE, LayoutType.FIGURE],
-    [LayoutType.TITLE, LayoutType.TEXT],
-    [LayoutType.TEXT, LayoutType.KEY_VALUE_AREA],
-    [LayoutType.TEXT, LayoutType.LIST_ITEM],
-    [LayoutType.TEXT, LayoutType.CAPTION],
-    [LayoutType.KEY_VALUE_AREA, LayoutType.LIST_ITEM],
-    [LayoutType.FIGURE, LayoutType.CAPTION],
+    [LayoutLabel.TABLE, LayoutLabel.TITLE],
+    [LayoutLabel.TABLE, LayoutLabel.TEXT],
+    [LayoutLabel.TABLE, LayoutLabel.KEY_VALUE_AREA],
+    [LayoutLabel.TABLE, LayoutLabel.LIST_ITEM],
+    [LayoutLabel.TABLE, LayoutLabel.LIST],
+    [LayoutLabel.TABLE, LayoutLabel.FIGURE],
+    [LayoutLabel.TITLE, LayoutLabel.TEXT],
+    [LayoutLabel.TEXT, LayoutLabel.KEY_VALUE_AREA],
+    [LayoutLabel.TEXT, LayoutLabel.LIST_ITEM],
+    [LayoutLabel.TEXT, LayoutLabel.CAPTION],
+    [LayoutLabel.KEY_VALUE_AREA, LayoutLabel.LIST_ITEM],
+    [LayoutLabel.FIGURE, LayoutLabel.CAPTION],
 ]
 cfg.LAYOUT_NMS_PAIRS.THRESHOLDS = [0.001, 0.01, 0.01, 0.001, 0.01, 0.01, 0.05, 0.01, 0.01, 0.01, 0.01, 0.001]
 cfg.LAYOUT_NMS_PAIRS.PRIORITY = [
-    LayoutType.TABLE,
-    LayoutType.TABLE,
-    LayoutType.TABLE,
-    LayoutType.TABLE,
-    LayoutType.TABLE,
-    LayoutType.TABLE,
-    LayoutType.TEXT,
-    LayoutType.TEXT,
+    LayoutLabel.TABLE,
+    LayoutLabel.TABLE,
+    LayoutLabel.TABLE,
+    LayoutLabel.TABLE,
+    LayoutLabel.TABLE,
+    LayoutLabel.TABLE,
+    LayoutLabel.TEXT,
+    LayoutLabel.TEXT,
     None,
-    LayoutType.CAPTION,
-    LayoutType.KEY_VALUE_AREA,
-    LayoutType.FIGURE,
+    LayoutLabel.CAPTION,
+    LayoutLabel.KEY_VALUE_AREA,
+    LayoutLabel.FIGURE,
 ]
 
 # Relevant when LIB = PT. Use either TorchScript weights via ITEM.WEIGHTS_TS
@@ -688,43 +688,43 @@ cfg.SEGMENTATION.STRETCH_RULE = "equal"
 
 # Specifies the layout category used to identify tables.
 # Used in both Deepdoctection and Table Transformer approaches.
-cfg.SEGMENTATION.TABLE_NAME = LayoutType.TABLE
+cfg.SEGMENTATION.TABLE_NAME = LayoutLabel.TABLE
 
 # Lists the layout or cell types used in the original Deepdoctection approach.
 # Used by TableSegmentationService for cell assignments.
-cfg.SEGMENTATION.CELL_NAMES = [CellType.COLUMN_HEADER, CellType.BODY, LayoutType.CELL]
+cfg.SEGMENTATION.CELL_NAMES = [CellLabel.COLUMN_HEADER, CellLabel.BODY, LayoutLabel.CELL]
 
 # Lists all cell types used by the Table Transformer approach (PubtablesSegmentationService).
 # LayoutType.CELL is synthetically generated and not predicted by the structure recognition model.
 cfg.SEGMENTATION.PUBTABLES_CELL_NAMES = [
-    LayoutType.CELL,
+    LayoutLabel.CELL,
 ]
 
 # Subset of PUBTABLES_CELL_NAMES that represent spanning/header cells.
 # These need to be matched with row or column elements.
 cfg.SEGMENTATION.PUBTABLES_SPANNING_CELL_NAMES = [
-    CellType.SPANNING,
+    CellLabel.SPANNING,
 ]
 
 # Lists the layout categories used to identify row and column elements.
 # Used by TableSegmentationService.
-cfg.SEGMENTATION.ITEM_NAMES = [LayoutType.ROW, LayoutType.COLUMN]
+cfg.SEGMENTATION.ITEM_NAMES = [LayoutLabel.ROW, LayoutLabel.COLUMN]
 
 # Equivalent to ITEM_NAMES but used in the Table Transformer approach.
-cfg.SEGMENTATION.PUBTABLES_ITEM_NAMES = [LayoutType.ROW, LayoutType.COLUMN]
+cfg.SEGMENTATION.PUBTABLES_ITEM_NAMES = [LayoutLabel.ROW, LayoutLabel.COLUMN]
 
 # Used in TableSegmentationService to specify sub-category annotations for row and column numbers.
-cfg.SEGMENTATION.SUB_ITEM_NAMES = [CellType.ROW_NUMBER, CellType.COLUMN_NUMBER]
+cfg.SEGMENTATION.SUB_ITEM_NAMES = [CellKey.ROW_NUMBER, CellKey.COLUMN_NUMBER]
 
 # Equivalent to SUB_ITEM_NAMES, but used with the Table Transformer approach.
-cfg.SEGMENTATION.PUBTABLES_SUB_ITEM_NAMES = [CellType.ROW_NUMBER, CellType.COLUMN_NUMBER]
+cfg.SEGMENTATION.PUBTABLES_SUB_ITEM_NAMES = [CellKey.ROW_NUMBER, CellKey.COLUMN_NUMBER]
 
 # Used in PubtablesSegmentationService.
 # Specifies which cells should be treated as header cells that need to be linked to row/column elements.
 cfg.SEGMENTATION.PUBTABLES_ITEM_HEADER_CELL_NAMES = [
-    CellType.COLUMN_HEADER,
-    CellType.ROW_HEADER,
-    CellType.PROJECTED_ROW_HEADER,
+    CellLabel.COLUMN_HEADER,
+    CellLabel.ROW_HEADER,
+    CellLabel.PROJECTED_ROW_HEADER,
 ]
 
 # Defines the threshold values for matching column/row header cells to their respective rows/columns
@@ -840,11 +840,11 @@ cfg.TEXT_ORDERING.PARAGRAPH_BREAK = 0.035
 
 # Specifies the parent layout categories in the link relationship.
 # These are the elements to which related components (e.g., captions) should be linked.
-cfg.LAYOUT_LINK.PARENTAL_CATEGORIES = [LayoutType.FIGURE, LayoutType.TABLE]
+cfg.LAYOUT_LINK.PARENTAL_CATEGORIES = [LayoutLabel.FIGURE, LayoutLabel.TABLE]
 
 # Specifies the child layout categories in the link relationship.
 # These are typically smaller or subordinate elements (e.g., captions).
-cfg.LAYOUT_LINK.CHILD_CATEGORIES = [LayoutType.CAPTION]
+cfg.LAYOUT_LINK.CHILD_CATEGORIES = [LayoutLabel.CAPTION]
 
 # Weights configuration for language detection model.
 cfg.LM_LANGUAGE_DETECT_CLASS.WEIGHTS = None

{deepdoctection-1.0.7 → deepdoctection-1.2.0}/src/deepdoctection/analyzer/factory.py

@@ -28,7 +28,7 @@ from lazy_imports import try_import
 from dd_core.utils.env_info import SETTINGS
 from dd_core.utils.error import DependencyError
 from dd_core.utils.metacfg import AttrDict
-from dd_core.utils.object_types import CellType, LayoutType, ObjectTypes, Relationships
+from dd_core.utils.object_types import CellLabel, LayoutLabel, ObjectTypes, RelationshipKey
 from dd_core.utils.transform import PadTransform
 
 from ..extern.base import ImageTransformer, ObjectDetector, PdfMiner
@@ -427,7 +427,7 @@ class ServiceFactory:
         if mode == "ITEM":
             if detector.__class__.__name__ in ("HFDetrDerivedDetector",):
                 exclude_category_names.extend(
-                    [LayoutType.TABLE, CellType.COLUMN_HEADER, CellType.PROJECTED_ROW_HEADER, CellType.SPANNING]
+                    [LayoutLabel.TABLE, CellLabel.COLUMN_HEADER, CellLabel.PROJECTED_ROW_HEADER, CellLabel.SPANNING]
                 )
         return {"exclude_category_names": exclude_category_names}
 
@@ -452,7 +452,7 @@ class ServiceFactory:
         )
         return SubImageLayoutService(
             sub_image_detector=detector,
-            sub_image_names=[LayoutType.TABLE, LayoutType.TABLE_ROTATED],
+            sub_image_names=[LayoutLabel.TABLE, LayoutLabel.TABLE_ROTATED],
             detect_result_generator=detect_result_generator,
             padder=padder,
         )
@@ -1031,14 +1031,14 @@ class ServiceFactory:
             FamilyCompound(
                 parent_categories=parental_categories,
                 child_categories=text_container,
-                relationship_key=Relationships.CHILD,
+                relationship_key=RelationshipKey.CHILD,
             ),
             FamilyCompound(
-                parent_categories=[LayoutType.LIST],
-                child_categories=[LayoutType.LIST_ITEM],
-                relationship_key=Relationships.CHILD,
+                parent_categories=[LayoutLabel.LIST],
+                child_categories=[LayoutLabel.LIST_ITEM],
+                relationship_key=RelationshipKey.CHILD,
                 create_synthetic_parent=True,
-                synthetic_parent=LayoutType.LIST,
+                synthetic_parent=LayoutLabel.LIST,
             ),
         ]
         return MatchingService(
@@ -1093,7 +1093,7 @@ class ServiceFactory:
             FamilyCompound(
                 parent_categories=parental_categories,
                 child_categories=child_categories,
-                relationship_key=Relationships.LAYOUT_LINK,
+                relationship_key=RelationshipKey.LAYOUT_LINK,
             )
         ]
         return MatchingService(
@@ -1153,9 +1153,9 @@ class ServiceFactory:
         )
         family_compounds = [
             FamilyCompound(
-                parent_categories=[LayoutType.LIST],
-                child_categories=[LayoutType.LINE],
-                relationship_key=Relationships.CHILD,
+                parent_categories=[LayoutLabel.LIST],
+                child_categories=[LayoutLabel.LINE],
+                relationship_key=RelationshipKey.CHILD,
             ),
         ]
         return MatchingService(
@@ -1612,7 +1612,7 @@ class ServiceFactory:
         token_classifier: Union[LayoutTokenModels, LmTokenModels],
         tokenizer_fast: Any,
         use_other_as_default_category: bool,
-        segment_positions: Union[LayoutType, Sequence[LayoutType], None],
+        segment_positions: Union[LayoutLabel, Sequence[LayoutLabel], None],
         sliding_window_stride: int,
     ) -> LMTokenClassifierService:
         """

{deepdoctection-1.0.7 → deepdoctection-1.2.0}/src/deepdoctection/eval/eval.py

@@ -34,7 +34,7 @@ from dd_core.mapper import filter_cat, remove_cats
 from dd_core.mapper.misc import maybe_load_image, maybe_remove_image, maybe_remove_image_from_category
 from dd_core.mapper.wandbstruct import to_wandb_image
 from dd_core.utils.logger import LoggingRecord, logger
-from dd_core.utils.object_types import DatasetType, LayoutType, TypeOrStr, get_type
+from dd_core.utils.object_types import DatasetKind, LayoutLabel, TypeOrStr, get_type
 from dd_core.utils.types import PixelValues
 from dd_core.utils.viz import interactive_imshow
 
@@ -146,14 +146,14 @@ class Evaluator:
 
         self.wandb_table_agent: Optional[WandbTableAgent]
         if run is not None:
-            if self.dataset.dataset_info.type == DatasetType.OBJECT_DETECTION:
+            if self.dataset.dataset_info.type == DatasetKind.OBJECT_DETECTION:
                 self.wandb_table_agent = WandbTableAgent(
                     run,
                     self.dataset.dataset_info.name,
                     50,
                     self.dataset.dataflow.categories.get_categories(filtered=True),
                 )
-            elif self.dataset.dataset_info.type == DatasetType.TOKEN_CLASSIFICATION:
+            elif self.dataset.dataset_info.type == DatasetKind.TOKEN_CLASSIFICATION:
                 if hasattr(self.metric, "sub_cats"):
                     sub_cat_key, sub_cat_val_list = list(self.metric.sub_cats.items())[0]
                     sub_cat_val = sub_cat_val_list[0]
@@ -253,7 +253,7 @@ class Evaluator:
         possible_cats_in_datapoint = self.dataset.dataflow.categories.get_categories(as_dict=False, filtered=True)
 
         # clean-up procedure depends on the dataset type
-        if self.dataset.dataset_info.type == DatasetType.OBJECT_DETECTION:
+        if self.dataset.dataset_info.type == DatasetKind.OBJECT_DETECTION:
             # we keep all image annotations that will not be generated through processing
             anns_to_keep = {ann for ann in possible_cats_in_datapoint if ann not in meta_anns.image_annotations}
             sub_cats_to_remove = meta_anns.sub_categories
@@ -269,11 +269,11 @@ class Evaluator:
                 remove_cats(sub_categories=sub_cats_to_remove, relationships=relationships_to_remove),
             )
 
-        elif self.dataset.dataset_info.type == DatasetType.SEQUENCE_CLASSIFICATION:
+        elif self.dataset.dataset_info.type == DatasetKind.SEQUENCE_CLASSIFICATION:
             summary_sub_cats_to_remove = meta_anns.summaries
             df_pr = MapData(df_pr, remove_cats(summary_sub_categories=summary_sub_cats_to_remove))
 
-        elif self.dataset.dataset_info.type == DatasetType.TOKEN_CLASSIFICATION:
+        elif self.dataset.dataset_info.type == DatasetKind.TOKEN_CLASSIFICATION:
             sub_cats_to_remove = meta_anns.sub_categories
             df_pr = MapData(df_pr, remove_cats(sub_categories=sub_cats_to_remove))
         else:
@@ -313,7 +313,7 @@ class Evaluator:
         df_pr = self._clean_up_predict_dataflow_annotations(df_pr)
 
         page_parsing_component = PageParsingService(
-            text_container=LayoutType.WORD,
+            text_container=LayoutLabel.WORD,
             floating_text_block_categories=floating_text_block_categories,  # type: ignore
             include_residual_text_container=bool(include_residual_text_containers),
         )

{deepdoctection-1.0.7 → deepdoctection-1.2.0}/src/deepdoctection/eval/tedsmetric.py

@@ -31,7 +31,7 @@ from dd_core.datapoint.image import Image
 from dd_core.datapoint.view import Page
 from dd_core.utils.file_utils import Requirement, get_apted_requirement, get_distance_requirement, get_lxml_requirement
 from dd_core.utils.logger import LoggingRecord, logger
-from dd_core.utils.object_types import LayoutType
+from dd_core.utils.object_types import LayoutLabel
 from dd_core.utils.types import MetricResults
 
 from .base import MetricBase
@@ -238,9 +238,9 @@ class TedsMetric(MetricBase):
     """
 
     metric = teds_metric
-    mapper: Callable[[Image, LayoutType, list[LayoutType]], Page] = Page.from_image
-    text_container: LayoutType = LayoutType.WORD
-    floating_text_block_categories = [LayoutType.TABLE]
+    mapper: Callable[[Image, LayoutLabel, list[LayoutLabel]], Page] = Page.from_image
+    text_container: LayoutLabel = LayoutLabel.WORD
+    floating_text_block_categories = [LayoutLabel.TABLE]
 
     structure_only = False
 

{deepdoctection-1.0.7 → deepdoctection-1.2.0}/src/deepdoctection/extern/deskew.py

@@ -24,7 +24,7 @@ from __future__ import annotations
 from lazy_imports import try_import
 
 from dd_core.utils.file_utils import get_jdeskew_requirement
-from dd_core.utils.object_types import ObjectTypes, PageType
+from dd_core.utils.object_types import ObjectTypes, PageKey
 from dd_core.utils.types import PixelValues, Requirement
 from dd_core.utils.viz import viz_handler
 
@@ -90,4 +90,4 @@ class Jdeskewer(ImageTransformer):
         return self.__class__(self.min_angle_rotation)
 
     def get_category_names(self) -> tuple[ObjectTypes, ...]:
-        return (PageType.ANGLE,)
+        return (PageKey.ANGLE,)

{deepdoctection-1.0.7 → deepdoctection-1.2.0}/src/deepdoctection/extern/doctrocr.py

@@ -35,7 +35,7 @@ from dd_core.utils.file_utils import (
     get_pytorch_requirement,
 )
 from dd_core.utils.fs import load_json
-from dd_core.utils.object_types import LayoutType, ObjectTypes, PageType, TypeOrStr
+from dd_core.utils.object_types import LayoutLabel, ObjectTypes, PageKey, TypeOrStr
 from dd_core.utils.transform import RotationTransform
 from dd_core.utils.types import PathLikeOrStr, PixelValues, Requirement
 from dd_core.utils.viz import viz_handler
@@ -84,7 +84,7 @@ def doctr_predict_text_lines(np_img: PixelValues, predictor: DetectionPredictor)
 
     detection_results = [
         DetectionResult(
-            box=box[:4].tolist(), class_id=1, score=box[4], absolute_coords=False, class_name=LayoutType.WORD
+            box=box[:4].tolist(), class_id=1, score=box[4], absolute_coords=False, class_name=LayoutLabel.WORD
         )
         for box in raw_output[0]["words"]
     ]
@@ -507,4 +507,4 @@ class DocTrRotationTransformer(ImageTransformer):
         return self.__class__(self.number_contours, self.ratio_threshold_for_lines)
 
     def get_category_names(self) -> tuple[ObjectTypes, ...]:
-        return (PageType.ANGLE,)
+        return (PageKey.ANGLE,)

{deepdoctection-1.0.7 → deepdoctection-1.2.0}/src/deepdoctection/extern/hfdetr.py

@@ -22,6 +22,7 @@ HF Detr and DeformableDetr models.
 from __future__ import annotations
 
 import os
+import warnings
 from abc import ABC
 from pathlib import Path
 from typing import TYPE_CHECKING, Literal, Mapping, Optional, Sequence, Union
@@ -41,6 +42,7 @@ with try_import() as pt_import_guard:
 
 with try_import() as tr_import_guard:
     from transformers import (
+        AutoConfig,
         DeformableDetrForObjectDetection,
         DeformableDetrImageProcessorFast,
         DetrImageProcessorFast,
@@ -265,14 +267,16 @@ class HFDetrDerivedDetector(HFDetrDerivedDetectorMixin):
         Raises:
             ValueError: If model architecture is not eligible.
         """
-        if "TableTransformerForObjectDetection" in config.architectures:
-            return TableTransformerForObjectDetection.from_pretrained(
-                pretrained_model_name_or_path=os.fspath(path_weights), config=config
-            )
-        if "DeformableDetrForObjectDetection" in config.architectures:
-            return DeformableDetrForObjectDetection.from_pretrained(
-                pretrained_model_name_or_path=os.fspath(path_weights), config=config
-            )
+        with warnings.catch_warnings():
+            warnings.filterwarnings("ignore", message=".*copying from a non-meta parameter.*")
+            if "TableTransformerForObjectDetection" in config.architectures:
+                return TableTransformerForObjectDetection.from_pretrained(
+                    pretrained_model_name_or_path=os.fspath(path_weights), config=config
+                )
+            if "DeformableDetrForObjectDetection" in config.architectures:
+                return DeformableDetrForObjectDetection.from_pretrained(
+                    pretrained_model_name_or_path=os.fspath(path_weights), config=config
+                )
         raise ValueError(
             f"Model architecture {config.architectures} not eligible. Please use either "
             "TableTransformerForObjectDetection or DeformableDetrForObjectDetection."
@@ -317,8 +321,14 @@ class HFDetrDerivedDetector(HFDetrDerivedDetectorMixin):
         Returns:
             `PretrainedConfig` instance.
         """
-        config = PretrainedConfig.from_pretrained(pretrained_model_name_or_path=os.fspath(path_config))
-        config.use_timm_backbone = True
+
+        config = AutoConfig.from_pretrained(pretrained_model_name_or_path=os.fspath(path_config))
+
+        # keep older behavior when supported by the concrete config
+        if hasattr(config, "use_timm_backbone"):
+            config.use_timm_backbone = True
+
+        # deepdoctection-specific runtime attributes
         config.threshold = 0.1
         config.nms_threshold = 0.05
         return config

{deepdoctection-1.0.7 → deepdoctection-1.2.0}/src/deepdoctection/extern/hflm.py

@@ -123,7 +123,7 @@ def predict_sequence_classes_from_lm(
 
     outputs = model(input_ids=input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids)
 
-    score = torch.max(F.softmax(outputs.logits)).tolist()
+    score = torch.max(F.softmax(outputs.logits, dim=1)).tolist()
     sequence_class_predictions = outputs.logits.argmax(-1).squeeze().tolist()
 
     return SequenceClassResult(class_id=sequence_class_predictions, score=float(score))

{deepdoctection-1.0.7 → deepdoctection-1.2.0}/src/deepdoctection/extern/pdftext.py

@@ -25,7 +25,7 @@ from lazy_imports import try_import
 
 from dd_core.utils.context import save_tmp_file
 from dd_core.utils.file_utils import get_pdfplumber_requirement, get_pypdfium2_requirement
-from dd_core.utils.object_types import LayoutType, ObjectTypes
+from dd_core.utils.object_types import LayoutLabel, ObjectTypes
 from dd_core.utils.types import Requirement
 
 from .base import DetectionResult, ModelCategories, PdfMiner
@@ -82,7 +82,7 @@ class PdfPlumberTextDetector(PdfMiner):
     def __init__(self, x_tolerance: int = 3, y_tolerance: int = 3) -> None:
         self.name = "Pdfplumber"
         self.model_id = self.get_model_id()
-        self.categories = ModelCategories(init_categories={1: LayoutType.WORD})
+        self.categories = ModelCategories(init_categories={1: LayoutLabel.WORD})
         self.x_tolerance = x_tolerance
         self.y_tolerance = y_tolerance
         self._page: Optional[Page] = None
@@ -169,7 +169,7 @@ class Pdfmium2TextDetector(PdfMiner):
     def __init__(self) -> None:
         self.name = "Pdfmium"
         self.model_id = self.get_model_id()
-        self.categories = ModelCategories(init_categories={1: LayoutType.LINE})
+        self.categories = ModelCategories(init_categories={1: LayoutLabel.LINE})
         self._page: Optional[Page] = None
 
     def predict(self, pdf_bytes: bytes) -> list[DetectionResult]:

{deepdoctection-1.0.7 → deepdoctection-1.2.0}/src/deepdoctection/extern/tessocr.py

@@ -37,7 +37,7 @@ from dd_core.utils.context import save_tmp_file, timeout_manager
 from dd_core.utils.error import DependencyError, TesseractError
 from dd_core.utils.file_utils import _TESS_PATH, get_tesseract_requirement
 from dd_core.utils.metacfg import config_to_cli_str, set_config_by_yaml
-from dd_core.utils.object_types import LayoutType, ObjectTypes, PageType
+from dd_core.utils.object_types import LayoutLabel, ObjectTypes, PageKey
 from dd_core.utils.transform import RotationTransform
 from dd_core.utils.types import PathLikeOrStr, PixelValues, Requirement
 from dd_core.utils.viz import viz_handler
@@ -248,7 +248,7 @@ def tesseract_line_to_detectresult(detect_result_list: list[DetectionResult]) ->
                 DetectionResult(
                     box=[ulx, uly, lrx, lry],
                     class_id=2,
-                    class_name=LayoutType.LINE,
+                    class_name=LayoutLabel.LINE,
                     text=" ".join(
                         [detect_result.text for detect_result in block_group if isinstance(detect_result.text, str)]
                     ),
@@ -295,7 +295,7 @@ def predict_text(np_img: PixelValues, supported_languages: str, text_lines: bool
                 score=score / 100,
                 text=caption[5],
                 class_id=1,
-                class_name=LayoutType.WORD,
+                class_name=LayoutLabel.WORD,
             )
             all_results.append(word)
     if text_lines:
@@ -381,9 +381,9 @@ class TesseractOcrDetector(ObjectDetector):
         self.config = hyper_param_config
 
         if self.config.LINES:
-            self.categories = ModelCategories(init_categories={1: LayoutType.WORD, 2: LayoutType.LINE})
+            self.categories = ModelCategories(init_categories={1: LayoutLabel.WORD, 2: LayoutLabel.LINE})
         else:
-            self.categories = ModelCategories(init_categories={1: LayoutType.WORD})
+            self.categories = ModelCategories(init_categories={1: LayoutLabel.WORD})
 
     def predict(self, np_img: PixelValues) -> list[DetectionResult]:
         """
@@ -455,7 +455,7 @@ class TesseractRotationTransformer(ImageTransformer):
 
     def __init__(self) -> None:
         self.name = fspath(_TESS_PATH) + "-rotation"
-        self.categories = ModelCategories(init_categories={1: PageType.ANGLE})
+        self.categories = ModelCategories(init_categories={1: PageKey.ANGLE})
         self.model_id = self.get_model_id()
         self.rotator = RotationTransform(360)
 

{deepdoctection-1.0.7 → deepdoctection-1.2.0}/src/deepdoctection/extern/texocr.py

@@ -28,7 +28,7 @@ from lazy_imports import try_import
 from dd_core.datapoint.convert import convert_np_array_to_b64_b
 from dd_core.utils.file_utils import get_boto3_requirement
 from dd_core.utils.logger import LoggingRecord, logger
-from dd_core.utils.object_types import LayoutType, ObjectTypes
+from dd_core.utils.object_types import LayoutLabel, ObjectTypes
 from dd_core.utils.types import JsonDict, PixelValues, Requirement
 
 from .base import DetectionResult, ModelCategories, ObjectDetector
@@ -54,7 +54,7 @@ def _textract_to_detectresult(response: JsonDict, width: int, height: int, text_
                     score=block["Confidence"] / 100,
                     text=block["Text"],
                     class_id=1 if block["BlockType"] == "WORD" else 2,
-                    class_name=LayoutType.WORD if block["BlockType"] == "WORD" else LayoutType.LINE,
+                    class_name=LayoutLabel.WORD if block["BlockType"] == "WORD" else LayoutLabel.LINE,
                 )
                 all_results.append(word)
 
@@ -142,9 +142,9 @@ class TextractOcrDetector(ObjectDetector):
         credentials_kwargs = self._maybe_resolve_secret(**credentials_kwargs)
         self.client = boto3.client("textract", **credentials_kwargs)
         if self.text_lines:
-            self.categories = ModelCategories(init_categories={1: LayoutType.WORD, 2: LayoutType.LINE})
+            self.categories = ModelCategories(init_categories={1: LayoutLabel.WORD, 2: LayoutLabel.LINE})
         else:
-            self.categories = ModelCategories(init_categories={1: LayoutType.WORD})
+            self.categories = ModelCategories(init_categories={1: LayoutLabel.WORD})
 
     def predict(self, np_img: PixelValues) -> list[DetectionResult]:
         """

{deepdoctection-1.0.7 → deepdoctection-1.2.0}/src/deepdoctection/pipe/anngen.py

@@ -28,7 +28,7 @@ from dd_core.datapoint.annotation import DEFAULT_CATEGORY_ID, CategoryAnnotation
 from dd_core.datapoint.box import BoundingBox, local_to_global_coords, rescale_coords
 from dd_core.datapoint.image import Image
 from dd_core.mapper.maputils import MappingContextManager
-from dd_core.utils.object_types import ObjectTypes, Relationships
+from dd_core.utils.object_types import ObjectTypes, RelationshipKey
 
 from ..extern.base import DetectionResult
 
@@ -121,6 +121,10 @@ class DatapointManager:
         self._cache_anns = {ann.annotation_id: ann for ann in dp.get_annotation()}
         self.datapoint_is_passed = True
 
+    def set_model_id(self, model_id: str | None) -> None:
+        """Re-sets the model_id."""
+        self.model_id = model_id
+
     def assert_datapoint_passed(self) -> None:
         """
         Asserts that a datapoint is passed.
@@ -227,7 +231,7 @@ class DatapointManager:
                     self.datapoint.image_id,
                     ann_global_box.transform(image_width=self.datapoint.width, image_height=self.datapoint.height),
                 )
-                parent_ann.dump_relationship(Relationships.CHILD, ann.annotation_id)
+                parent_ann.dump_relationship(RelationshipKey.CHILD, ann.annotation_id)
 
             self.datapoint.dump(ann)
             self._cache_anns[ann.annotation_id] = ann

{deepdoctection-1.0.7 → deepdoctection-1.2.0}/src/deepdoctection/pipe/base.py

@@ -63,7 +63,7 @@ class PipelineComponent(ABC):
         Currently, predictors can only process single images. Processing higher number of batches is not planned.
     """
 
-    def __init__(self, name: str, model_id: Optional[str] = None) -> None:
+    def __init__(self, name: str, model_id: Optional[str] = None, service_id: Optional[str] = None) -> None:
         """
         Initializes a `PipelineComponent`.
 
@@ -71,9 +71,10 @@ class PipelineComponent(ABC):
             name: The name of the pipeline component. The name will be used to identify a pipeline component in a
                   pipeline. Use something that describes the task of the pipeline.
             model_id: Optional model identifier.
+            service_id: Optional service identifier override to avoid name collisions.
         """
         self.name = name
-        self.service_id = self.get_service_id()
+        self.service_id = service_id or self.get_service_id()
         self.dp_manager = DatapointManager(self.service_id, model_id)
         self.timer_on = False
         self.filter_func: Callable[[DP], bool] = lambda dp: False