mirrors/docling
1
0
Fork 0
mirror of https://github.com/DS4SD/docling.git synced 2025-07-29 21:44:32 +00:00
Code Issues Packages Projects Releases Wiki Activity

word-level pdf cells for tables

Browse Source 
Signed-off-by: Maksym Lysak <mly@zurich.ibm.com>
This commit is contained in:
Maksym Lysak 2025-03-25 13:36:58 +01:00
parent 82694b2136
commit 787c6d8ace
1 changed files with 29 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

33
docling/models/table_structure_model.py
View File

@ -5,7 +5,11 @@ from typing import Iterable, Optional, Union
import numpy
from docling_core.types.doc import BoundingBox, DocItemLabel, TableCell
from docling_core.types.doc.page import BoundingRectangle
from docling_core.types.doc.page import (
BoundingRectangle,
SegmentedPdfPage,
TextCellUnit,
)
from docling_ibm_models.tableformer.data_management.tf_predictor import TFPredictor
from PIL import ImageDraw
@ -218,9 +222,30 @@ class TableStructureModel(BasePageModel):
if len(table_bboxes):
for table_cluster, tbl_box in in_tables:
# Check if word-level cells are available from backend:
sp = page._backend.get_segmented_page()
if sp is not None:
tcells = sp.get_cells_in_bbox(
cell_unit=TextCellUnit.WORD,
bbox=table_cluster.bbox.to_bottom_left_origin(
page.size.height
),
)
# Transform origin of returned cells:
tcells2 = []
for tcell in tcells:
# Do the copy to not affect cells that are in two (or more) table regions
tcell = tcell.model_copy(deep=True)
tcell.rect = tcell.rect.to_top_left_origin(
page.size.height
)
tcells2.append(tcell)
tcells = tcells2
else:
print("Otherwise - we use normal (line/phrase) cells")
tcells = table_cluster.cells
tokens = []
for c in table_cluster.cells:
for c in tcells:
# Only allow non empty stings (spaces) into the cells of a table
if len(c.text.strip()) > 0:
new_cell = copy.deepcopy(c)
@ -229,7 +254,6 @@ class TableStructureModel(BasePageModel):
scale=self.scale
)
)
tokens.append(
{
"id": new_cell.index,
@ -244,6 +268,7 @@ class TableStructureModel(BasePageModel):
)
table_out = tf_output[0]
table_cells = []
# print("len(tf_responses)={}".format(len(table_out["tf_responses"])))
for element in table_out["tf_responses"]:
if not self.do_cell_matching: