docling 1.19.1__py3-none-any.whl → 2.1.0__py3-none-any.whl

docling/models/layout_model.py

@@ -2,8 +2,10 @@ import copy
 import logging
 import random
 import time
+from pathlib import Path
 from typing import Iterable, List
 
+from docling_core.types.doc import CoordOrigin, DocItemLabel
 from docling_ibm_models.layoutmodel.layout_predictor import LayoutPredictor
 from PIL import ImageDraw
 
@@ -11,74 +13,73 @@ from docling.datamodel.base_models import (
     BoundingBox,
     Cell,
     Cluster,
-    CoordOrigin,
     LayoutPrediction,
     Page,
 )
+from docling.models.base_model import BasePageModel
 from docling.utils import layout_utils as lu
 
 _log = logging.getLogger(__name__)
 
 
-class LayoutModel:
+class LayoutModel(BasePageModel):
 
     TEXT_ELEM_LABELS = [
-        "Text",
-        "Footnote",
-        "Caption",
-        "Checkbox-Unselected",
-        "Checkbox-Selected",
-        "Section-header",
-        "Page-header",
-        "Page-footer",
-        "Code",
-        "List-item",
-        # "Title"
+        DocItemLabel.TEXT,
+        DocItemLabel.FOOTNOTE,
+        DocItemLabel.CAPTION,
+        DocItemLabel.CHECKBOX_UNSELECTED,
+        DocItemLabel.CHECKBOX_SELECTED,
+        DocItemLabel.SECTION_HEADER,
+        DocItemLabel.PAGE_HEADER,
+        DocItemLabel.PAGE_FOOTER,
+        DocItemLabel.CODE,
+        DocItemLabel.LIST_ITEM,
         # "Formula",
     ]
-    PAGE_HEADER_LABELS = ["Page-header", "Page-footer"]
+    PAGE_HEADER_LABELS = [DocItemLabel.PAGE_HEADER, DocItemLabel.PAGE_FOOTER]
 
-    TABLE_LABEL = "Table"
-    FIGURE_LABEL = "Picture"
-    FORMULA_LABEL = "Formula"
+    TABLE_LABEL = DocItemLabel.TABLE
+    FIGURE_LABEL = DocItemLabel.PICTURE
+    FORMULA_LABEL = DocItemLabel.FORMULA
 
-    def __init__(self, config):
-        self.config = config
-        self.layout_predictor = LayoutPredictor(
-            config["artifacts_path"]
-        )  # TODO temporary
+    def __init__(self, artifacts_path: Path):
+        self.layout_predictor = LayoutPredictor(artifacts_path)  # TODO temporary
 
-    def postprocess(self, clusters: List[Cluster], cells: List[Cell], page_height):
+    def postprocess(self, clusters_in: List[Cluster], cells: List[Cell], page_height):
         MIN_INTERSECTION = 0.2
         CLASS_THRESHOLDS = {
-            "Caption": 0.35,
-            "Footnote": 0.35,
-            "Formula": 0.35,
-            "List-item": 0.35,
-            "Page-footer": 0.35,
-            "Page-header": 0.35,
-            "Picture": 0.2,  # low threshold adjust to capture chemical structures for examples.
-            "Section-header": 0.45,
-            "Table": 0.35,
-            "Text": 0.45,
-            "Title": 0.45,
-            "Document Index": 0.45,
-            "Code": 0.45,
-            "Checkbox-Selected": 0.45,
-            "Checkbox-Unselected": 0.45,
-            "Form": 0.45,
-            "Key-Value Region": 0.45,
+            DocItemLabel.CAPTION: 0.35,
+            DocItemLabel.FOOTNOTE: 0.35,
+            DocItemLabel.FORMULA: 0.35,
+            DocItemLabel.LIST_ITEM: 0.35,
+            DocItemLabel.PAGE_FOOTER: 0.35,
+            DocItemLabel.PAGE_HEADER: 0.35,
+            DocItemLabel.PICTURE: 0.2,  # low threshold adjust to capture chemical structures for examples.
+            DocItemLabel.SECTION_HEADER: 0.45,
+            DocItemLabel.TABLE: 0.35,
+            DocItemLabel.TEXT: 0.45,
+            DocItemLabel.TITLE: 0.45,
+            DocItemLabel.DOCUMENT_INDEX: 0.45,
+            DocItemLabel.CODE: 0.45,
+            DocItemLabel.CHECKBOX_SELECTED: 0.45,
+            DocItemLabel.CHECKBOX_UNSELECTED: 0.45,
+            DocItemLabel.FORM: 0.45,
+            DocItemLabel.KEY_VALUE_REGION: 0.45,
         }
 
-        CLASS_REMAPPINGS = {"Document Index": "Table", "Title": "Section-header"}
+        CLASS_REMAPPINGS = {
+            DocItemLabel.DOCUMENT_INDEX: DocItemLabel.TABLE,
+            DocItemLabel.TITLE: DocItemLabel.SECTION_HEADER,
+        }
 
         _log.debug("================= Start postprocess function ====================")
         start_time = time.time()
         # Apply Confidence Threshold to cluster predictions
         # confidence = self.conf_threshold
-        clusters_out = []
+        clusters_mod = []
 
-        for cluster in clusters:
+        for cluster in clusters_in:
             confidence = CLASS_THRESHOLDS[cluster.label]
             if cluster.confidence >= confidence:
                 # annotation["created_by"] = "high_conf_pred"
@@ -86,10 +87,10 @@ class LayoutModel:
                 # Remap class labels where needed.
                 if cluster.label in CLASS_REMAPPINGS.keys():
                     cluster.label = CLASS_REMAPPINGS[cluster.label]
-                clusters_out.append(cluster)
+                clusters_mod.append(cluster)
 
         # map to dictionary clusters and cells, with bottom left origin
-        clusters = [
+        clusters_orig = [
             {
                 "id": c.id,
                 "bbox": list(
@@ -99,7 +100,7 @@ class LayoutModel:
                 "cell_ids": [],
                 "type": c.label,
             }
-            for c in clusters
+            for c in clusters_in
         ]
 
         clusters_out = [
@@ -113,9 +114,11 @@ class LayoutModel:
                 "cell_ids": [],
                 "type": c.label,
             }
-            for c in clusters_out
+            for c in clusters_mod
         ]
 
+        del clusters_mod
+
         raw_cells = [
             {
                 "id": c.id,
@@ -149,7 +152,7 @@ class LayoutModel:
 
         # Assign orphan cells with lower confidence predictions
         clusters_out, orphan_cell_indices = lu.assign_orphans_with_low_conf_pred(
-            clusters_out, clusters, raw_cells, orphan_cell_indices
+            clusters_out, clusters_orig, raw_cells, orphan_cell_indices
         )
 
         # Refresh the cell_ids assignment, after creating new clusters using low conf predictions
@@ -178,7 +181,7 @@ class LayoutModel:
         ) = lu.cell_id_state_map(clusters_out, cell_count)
 
         clusters_out, orphan_cell_indices = lu.set_orphan_as_text(
-            clusters_out, clusters, raw_cells, orphan_cell_indices
+            clusters_out, clusters_orig, raw_cells, orphan_cell_indices
         )
 
         _log.debug("---- 5. Merge Cells & and adapt the bounding boxes")
@@ -237,92 +240,105 @@ class LayoutModel:
         end_time = time.time() - start_time
         _log.debug(f"Finished post processing in seconds={end_time:.3f}")
 
-        cells_out = [
+        cells_out_new = [
             Cell(
-                id=c["id"],
+                id=c["id"],  # type: ignore
                 bbox=BoundingBox.from_tuple(
-                    coord=c["bbox"], origin=CoordOrigin.BOTTOMLEFT
+                    coord=c["bbox"], origin=CoordOrigin.BOTTOMLEFT  # type: ignore
                 ).to_top_left_origin(page_height),
-                text=c["text"],
+                text=c["text"],  # type: ignore
             )
             for c in cells_out
         ]
+
+        del cells_out
+
         clusters_out_new = []
         for c in clusters_out:
-            cluster_cells = [ccell for ccell in cells_out if ccell.id in c["cell_ids"]]
+            cluster_cells = [
+                ccell for ccell in cells_out_new if ccell.id in c["cell_ids"]  # type: ignore
+            ]
             c_new = Cluster(
-                id=c["id"],
+                id=c["id"],  # type: ignore
                 bbox=BoundingBox.from_tuple(
-                    coord=c["bbox"], origin=CoordOrigin.BOTTOMLEFT
+                    coord=c["bbox"], origin=CoordOrigin.BOTTOMLEFT  # type: ignore
                 ).to_top_left_origin(page_height),
-                confidence=c["confidence"],
-                label=c["type"],
+                confidence=c["confidence"],  # type: ignore
+                label=DocItemLabel(c["type"]),
                 cells=cluster_cells,
             )
             clusters_out_new.append(c_new)
 
-        return clusters_out_new, cells_out
+        return clusters_out_new, cells_out_new
 
     def __call__(self, page_batch: Iterable[Page]) -> Iterable[Page]:
         for page in page_batch:
-            clusters = []
-            for ix, pred_item in enumerate(
-                self.layout_predictor.predict(page.get_image(scale=1.0))
-            ):
-                cluster = Cluster(
-                    id=ix,
-                    label=pred_item["label"],
-                    confidence=pred_item["confidence"],
-                    bbox=BoundingBox.model_validate(pred_item),
-                    cells=[],
-                )
-
-                clusters.append(cluster)
-
-            # Map cells to clusters
-            # TODO: Remove, postprocess should take care of it anyway.
-            for cell in page.cells:
-                for cluster in clusters:
-                    if not cell.bbox.area() > 0:
-                        overlap_frac = 0.0
-                    else:
-                        overlap_frac = (
-                            cell.bbox.intersection_area_with(cluster.bbox)
-                            / cell.bbox.area()
-                        )
-
-                    if overlap_frac > 0.5:
-                        cluster.cells.append(cell)
-
-            # Pre-sort clusters
-            # clusters = self.sort_clusters_by_cell_order(clusters)
-
-            # DEBUG code:
-            def draw_clusters_and_cells():
-                image = copy.deepcopy(page.image)
-                draw = ImageDraw.Draw(image)
-                for c in clusters:
-                    x0, y0, x1, y1 = c.bbox.as_tuple()
-                    draw.rectangle([(x0, y0), (x1, y1)], outline="green")
-
-                    cell_color = (
-                        random.randint(30, 140),
-                        random.randint(30, 140),
-                        random.randint(30, 140),
+            assert page._backend is not None
+            if not page._backend.is_valid():
+                yield page
+            else:
+                assert page.size is not None
+
+                clusters = []
+                for ix, pred_item in enumerate(
+                    self.layout_predictor.predict(page.get_image(scale=1.0))
+                ):
+                    label = DocItemLabel(
+                        pred_item["label"].lower().replace(" ", "_").replace("-", "_")
+                    )  # Temporary, until docling-ibm-model uses docling-core types
+                    cluster = Cluster(
+                        id=ix,
+                        label=label,
+                        confidence=pred_item["confidence"],
+                        bbox=BoundingBox.model_validate(pred_item),
+                        cells=[],
                     )
-                    for tc in c.cells:  # [:1]:
-                        x0, y0, x1, y1 = tc.bbox.as_tuple()
-                        draw.rectangle([(x0, y0), (x1, y1)], outline=cell_color)
-                image.show()
+                    clusters.append(cluster)
+
+                # Map cells to clusters
+                # TODO: Remove, postprocess should take care of it anyway.
+                for cell in page.cells:
+                    for cluster in clusters:
+                        if not cell.bbox.area() > 0:
+                            overlap_frac = 0.0
+                        else:
+                            overlap_frac = (
+                                cell.bbox.intersection_area_with(cluster.bbox)
+                                / cell.bbox.area()
+                            )
+
+                        if overlap_frac > 0.5:
+                            cluster.cells.append(cell)
+
+                # Pre-sort clusters
+                # clusters = self.sort_clusters_by_cell_order(clusters)
+
+                # DEBUG code:
+                def draw_clusters_and_cells():
+                    image = copy.deepcopy(page.image)
+                    draw = ImageDraw.Draw(image)
+                    for c in clusters:
+                        x0, y0, x1, y1 = c.bbox.as_tuple()
+                        draw.rectangle([(x0, y0), (x1, y1)], outline="green")
+
+                        cell_color = (
+                            random.randint(30, 140),
+                            random.randint(30, 140),
+                            random.randint(30, 140),
+                        )
+                        for tc in c.cells:  # [:1]:
+                            x0, y0, x1, y1 = tc.bbox.as_tuple()
+                            draw.rectangle([(x0, y0), (x1, y1)], outline=cell_color)
+                    image.show()
 
-            # draw_clusters_and_cells()
+                # draw_clusters_and_cells()
 
-            clusters, page.cells = self.postprocess(
-                clusters, page.cells, page.size.height
-            )
+                clusters, page.cells = self.postprocess(
+                    clusters, page.cells, page.size.height
+                )
 
-            # draw_clusters_and_cells()
+                # draw_clusters_and_cells()
 
-            page.predictions.layout = LayoutPrediction(clusters=clusters)
+                page.predictions.layout = LayoutPrediction(clusters=clusters)
 
-            yield page
+                yield page

docling/models/page_assemble_model.py

@@ -2,22 +2,29 @@ import logging
 import re
 from typing import Iterable, List
 
+from pydantic import BaseModel
+
 from docling.datamodel.base_models import (
     AssembledUnit,
     FigureElement,
     Page,
     PageElement,
-    TableElement,
+    Table,
     TextElement,
 )
+from docling.models.base_model import BasePageModel
 from docling.models.layout_model import LayoutModel
 
 _log = logging.getLogger(__name__)
 
 
-class PageAssembleModel:
-    def __init__(self, config):
-        self.config = config
+class PageAssembleOptions(BaseModel):
+    keep_images: bool = False
+
+
+class PageAssembleModel(BasePageModel):
+    def __init__(self, options: PageAssembleOptions):
+        self.options = options
 
     def sanitize_text(self, lines):
         if len(lines) <= 1:
@@ -46,103 +53,120 @@ class PageAssembleModel:
 
     def __call__(self, page_batch: Iterable[Page]) -> Iterable[Page]:
         for page in page_batch:
-            # assembles some JSON output page by page.
-
-            elements: List[PageElement] = []
-            headers: List[PageElement] = []
-            body: List[PageElement] = []
-
-            for cluster in page.predictions.layout.clusters:
-                # _log.info("Cluster label seen:", cluster.label)
-                if cluster.label in LayoutModel.TEXT_ELEM_LABELS:
-
-                    textlines = [
-                        cell.text.replace("\x02", "-").strip()
-                        for cell in cluster.cells
-                        if len(cell.text.strip()) > 0
-                    ]
-                    text = self.sanitize_text(textlines)
-                    text_el = TextElement(
-                        label=cluster.label,
-                        id=cluster.id,
-                        text=text,
-                        page_no=page.page_no,
-                        cluster=cluster,
-                    )
-                    elements.append(text_el)
-
-                    if cluster.label in LayoutModel.PAGE_HEADER_LABELS:
-                        headers.append(text_el)
-                    else:
-                        body.append(text_el)
-                elif cluster.label == LayoutModel.TABLE_LABEL:
-                    tbl = None
-                    if page.predictions.tablestructure:
-                        tbl = page.predictions.tablestructure.table_map.get(
-                            cluster.id, None
-                        )
-                    if (
-                        not tbl
-                    ):  # fallback: add table without structure, if it isn't present
-                        tbl = TableElement(
-                            label=cluster.label,
-                            id=cluster.id,
-                            text="",
-                            otsl_seq=[],
-                            table_cells=[],
-                            cluster=cluster,
-                            page_no=page.page_no,
-                        )
+            assert page._backend is not None
+            if not page._backend.is_valid():
+                yield page
+            else:
+                assert page.predictions.layout is not None
 
-                    elements.append(tbl)
-                    body.append(tbl)
-                elif cluster.label == LayoutModel.FIGURE_LABEL:
-                    fig = None
-                    if page.predictions.figures_classification:
-                        fig = page.predictions.figures_classification.figure_map.get(
-                            cluster.id, None
-                        )
-                    if (
-                        not fig
-                    ):  # fallback: add figure without classification, if it isn't present
-                        fig = FigureElement(
+                # assembles some JSON output page by page.
+
+                elements: List[PageElement] = []
+                headers: List[PageElement] = []
+                body: List[PageElement] = []
+
+                for cluster in page.predictions.layout.clusters:
+                    # _log.info("Cluster label seen:", cluster.label)
+                    if cluster.label in LayoutModel.TEXT_ELEM_LABELS:
+
+                        textlines = [
+                            cell.text.replace("\x02", "-").strip()
+                            for cell in cluster.cells
+                            if len(cell.text.strip()) > 0
+                        ]
+                        text = self.sanitize_text(textlines)
+                        text_el = TextElement(
                             label=cluster.label,
                             id=cluster.id,
-                            text="",
-                            data=None,
-                            cluster=cluster,
+                            text=text,
                             page_no=page.page_no,
+                            cluster=cluster,
                         )
-                    elements.append(fig)
-                    body.append(fig)
-                elif cluster.label == LayoutModel.FORMULA_LABEL:
-                    equation = None
-                    if page.predictions.equations_prediction:
-                        equation = (
-                            page.predictions.equations_prediction.equation_map.get(
+                        elements.append(text_el)
+
+                        if cluster.label in LayoutModel.PAGE_HEADER_LABELS:
+                            headers.append(text_el)
+                        else:
+                            body.append(text_el)
+                    elif cluster.label == LayoutModel.TABLE_LABEL:
+                        tbl = None
+                        if page.predictions.tablestructure:
+                            tbl = page.predictions.tablestructure.table_map.get(
                                 cluster.id, None
                             )
-                        )
-                    if not equation:  # fallback: add empty formula, if it isn't present
-                        text = self.sanitize_text(
-                            [
-                                cell.text.replace("\x02", "-").strip()
-                                for cell in cluster.cells
-                                if len(cell.text.strip()) > 0
-                            ]
-                        )
-                        equation = TextElement(
-                            label=cluster.label,
-                            id=cluster.id,
-                            cluster=cluster,
-                            page_no=page.page_no,
-                            text=text,
-                        )
-                    elements.append(equation)
-                    body.append(equation)
+                        if (
+                            not tbl
+                        ):  # fallback: add table without structure, if it isn't present
+                            tbl = Table(
+                                label=cluster.label,
+                                id=cluster.id,
+                                text="",
+                                otsl_seq=[],
+                                table_cells=[],
+                                cluster=cluster,
+                                page_no=page.page_no,
+                            )
+
+                        elements.append(tbl)
+                        body.append(tbl)
+                    elif cluster.label == LayoutModel.FIGURE_LABEL:
+                        fig = None
+                        if page.predictions.figures_classification:
+                            fig = (
+                                page.predictions.figures_classification.figure_map.get(
+                                    cluster.id, None
+                                )
+                            )
+                        if (
+                            not fig
+                        ):  # fallback: add figure without classification, if it isn't present
+                            fig = FigureElement(
+                                label=cluster.label,
+                                id=cluster.id,
+                                text="",
+                                data=None,
+                                cluster=cluster,
+                                page_no=page.page_no,
+                            )
+                        elements.append(fig)
+                        body.append(fig)
+                    elif cluster.label == LayoutModel.FORMULA_LABEL:
+                        equation = None
+                        if page.predictions.equations_prediction:
+                            equation = (
+                                page.predictions.equations_prediction.equation_map.get(
+                                    cluster.id, None
+                                )
+                            )
+                        if (
+                            not equation
+                        ):  # fallback: add empty formula, if it isn't present
+                            text = self.sanitize_text(
+                                [
+                                    cell.text.replace("\x02", "-").strip()
+                                    for cell in cluster.cells
+                                    if len(cell.text.strip()) > 0
+                                ]
+                            )
+                            equation = TextElement(
+                                label=cluster.label,
+                                id=cluster.id,
+                                cluster=cluster,
+                                page_no=page.page_no,
+                                text=text,
+                            )
+                        elements.append(equation)
+                        body.append(equation)
+
+                page.assembled = AssembledUnit(
+                    elements=elements, headers=headers, body=body
+                )
+
+                # Remove page images (can be disabled)
+                if not self.options.keep_images:
+                    page._image_cache = {}
 
-            page.assembled = AssembledUnit(
-                elements=elements, headers=headers, body=body
-            )
+                # Unload backend
+                page._backend.unload()
 
-            yield page
+                yield page

docling/models/page_preprocessing_model.py

@@ -0,0 +1,61 @@
+from typing import Iterable, Optional
+
+from PIL import ImageDraw
+from pydantic import BaseModel
+
+from docling.datamodel.base_models import Page
+from docling.models.base_model import BasePageModel
+
+
+class PagePreprocessingOptions(BaseModel):
+    images_scale: Optional[float]
+
+
+class PagePreprocessingModel(BasePageModel):
+    def __init__(self, options: PagePreprocessingOptions):
+        self.options = options
+
+    def __call__(self, page_batch: Iterable[Page]) -> Iterable[Page]:
+        for page in page_batch:
+            assert page._backend is not None
+            if not page._backend.is_valid():
+                yield page
+            else:
+                page = self._populate_page_images(page)
+                page = self._parse_page_cells(page)
+                yield page
+
+    # Generate the page image and store it in the page object
+    def _populate_page_images(self, page: Page) -> Page:
+        # default scale
+        page.get_image(
+            scale=1.0
+        )  # puts the page image on the image cache at default scale
+
+        images_scale = self.options.images_scale
+        # user requested scales
+        if images_scale is not None:
+            page._default_image_scale = images_scale
+            page.get_image(
+                scale=images_scale
+            )  # this will trigger storing the image in the internal cache
+
+        return page
+
+    # Extract and populate the page cells and store it in the page object
+    def _parse_page_cells(self, page: Page) -> Page:
+        assert page._backend is not None
+
+        page.cells = list(page._backend.get_text_cells())
+
+        # DEBUG code:
+        def draw_text_boxes(image, cells):
+            draw = ImageDraw.Draw(image)
+            for c in cells:
+                x0, y0, x1, y1 = c.bbox.as_tuple()
+                draw.rectangle([(x0, y0), (x1, y1)], outline="red")
+            image.show()
+
+        # draw_text_boxes(page.get_image(scale=1.0), cells)
+
+        return page