NeoAnd/Docling
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: Make Page.parsed_page the only source of truth for text cells, add OCR cells to it (#1745)

Browse Source 
* Keep page.parsed_page.textline_cells and page.cells in sync, including OCR

Signed-off-by: Christoph Auer <cau@zurich.ibm.com>

* Make page.parsed_page the only source of truth for text cells

Signed-off-by: Christoph Auer <cau@zurich.ibm.com>

* Small fix

Signed-off-by: Christoph Auer <cau@zurich.ibm.com>

* Correctly compute PDF boxes from pymupdf

Signed-off-by: Christoph Auer <cau@zurich.ibm.com>

* Use different OCR engine order

Signed-off-by: Christoph Auer <cau@zurich.ibm.com>

* Add type hints and fix mypy

Signed-off-by: Christoph Auer <cau@zurich.ibm.com>

* One more test fix

Signed-off-by: Christoph Auer <cau@zurich.ibm.com>

* Remove with pypdfium2_lock from caller sites

Signed-off-by: Christoph Auer <cau@zurich.ibm.com>

* Fix typing

Signed-off-by: Christoph Auer <cau@zurich.ibm.com>

---------

Signed-off-by: Christoph Auer <cau@zurich.ibm.com>
This commit is contained in:
Christoph Auer
2025-06-13 19:01:55 +02:00
committed by GitHub
parent 0432a31b2f
commit 7d3302cb48
50 changed files with 339091 additions and 330047 deletions
Download Patch File Download Diff File Expand all files Collapse all files

138
docling/backend/docling_parse_backend.py
View File

@@ -7,12 +7,17 @@ from typing import List, Optional, Union
import pypdfium2 as pdfium
from docling_core.types.doc import BoundingBox, CoordOrigin, Size
from docling_core.types.doc.page import BoundingRectangle, SegmentedPdfPage, TextCell
from docling_core.types.doc.page import (
BoundingRectangle,
SegmentedPdfPage,
TextCell,
)
from docling_parse.pdf_parsers import pdf_parser_v1
from PIL import Image, ImageDraw
from pypdfium2 import PdfPage
from docling.backend.pdf_backend import PdfDocumentBackend, PdfPageBackend
from docling.backend.pypdfium2_backend import get_pdf_page_geometry
from docling.datamodel.document import InputDocument
_log = logging.getLogger(__name__)
@@ -36,6 +41,51 @@ class DoclingParsePageBackend(PdfPageBackend):
def is_valid(self) -> bool:
return self.valid
def _compute_text_cells(self) -> List[TextCell]:
"""Compute text cells from docling-parse data."""
cells: List[TextCell] = []
cell_counter = 0
if not self.valid:
return cells
page_size = self.get_size()
parser_width = self._dpage["width"]
parser_height = self._dpage["height"]
for i in range(len(self._dpage["cells"])):
rect = self._dpage["cells"][i]["box"]["device"]
x0, y0, x1, y1 = rect
if x1 < x0:
x0, x1 = x1, x0
if y1 < y0:
y0, y1 = y1, y0
text_piece = self._dpage["cells"][i]["content"]["rnormalized"]
cells.append(
TextCell(
index=cell_counter,
text=text_piece,
orig=text_piece,
from_ocr=False,
rect=BoundingRectangle.from_bounding_box(
BoundingBox(
l=x0 * page_size.width / parser_width,
b=y0 * page_size.height / parser_height,
r=x1 * page_size.width / parser_width,
t=y1 * page_size.height / parser_height,
coord_origin=CoordOrigin.BOTTOMLEFT,
)
).to_top_left_origin(page_size.height),
)
)
cell_counter += 1
return cells
def get_text_in_rect(self, bbox: BoundingBox) -> str:
if not self.valid:
return ""
@@ -70,75 +120,27 @@ class DoclingParsePageBackend(PdfPageBackend):
return text_piece
def get_segmented_page(self) -> Optional[SegmentedPdfPage]:
return None
if not self.valid:
return None
text_cells = self._compute_text_cells()
# Get the PDF page geometry from pypdfium2
dimension = get_pdf_page_geometry(self._ppage)
# Create SegmentedPdfPage
return SegmentedPdfPage(
dimension=dimension,
textline_cells=text_cells,
char_cells=[],
word_cells=[],
has_lines=len(text_cells) > 0,
has_words=False,
has_chars=False,
)
def get_text_cells(self) -> Iterable[TextCell]:
cells: List[TextCell] = []
cell_counter = 0
if not self.valid:
return cells
page_size = self.get_size()
parser_width = self._dpage["width"]
parser_height = self._dpage["height"]
for i in range(len(self._dpage["cells"])):
rect = self._dpage["cells"][i]["box"]["device"]
x0, y0, x1, y1 = rect
if x1 < x0:
x0, x1 = x1, x0
if y1 < y0:
y0, y1 = y1, y0
text_piece = self._dpage["cells"][i]["content"]["rnormalized"]
cells.append(
TextCell(
index=cell_counter,
text=text_piece,
orig=text_piece,
from_ocr=False,
rect=BoundingRectangle.from_bounding_box(
BoundingBox(
# l=x0, b=y0, r=x1, t=y1,
l=x0 * page_size.width / parser_width,
b=y0 * page_size.height / parser_height,
r=x1 * page_size.width / parser_width,
t=y1 * page_size.height / parser_height,
coord_origin=CoordOrigin.BOTTOMLEFT,
)
).to_top_left_origin(page_size.height),
)
)
cell_counter += 1
def draw_clusters_and_cells():
image = (
self.get_page_image()
) # make new image to avoid drawing on the saved ones
draw = ImageDraw.Draw(image)
for c in cells:
x0, y0, x1, y1 = c.rect.to_bounding_box().as_tuple()
cell_color = (
random.randint(30, 140),
random.randint(30, 140),
random.randint(30, 140),
)
draw.rectangle([(x0, y0), (x1, y1)], outline=cell_color)
image.show()
# before merge:
# draw_clusters_and_cells()
# cells = merge_horizontal_cells(cells)
# after merge:
# draw_clusters_and_cells()
return cells
return self._compute_text_cells()
def get_bitmap_rects(self, scale: float = 1) -> Iterable[BoundingBox]:
AREA_THRESHOLD = 0 # 32 * 32