mirrors/docling
1
0
Fork 0
mirror of https://github.com/DS4SD/docling.git synced 2025-12-10 13:48:13 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix: Test cases for RTL programmatic PDFs and fixes for the formula model (#903)

Browse Source 
fix: Support for RTL programmatic documents
fix(parser): detect and handle rotated pages
fix(parser): fix bug causing duplicated text
fix(formula): improve stopping criteria
chore: update lock file
fix: temporary constrain beautifulsoup


* switch to code formula model v1.0.1 and new test pdf

Signed-off-by: Matteo-Omenetti <Matteo.Omenetti1@ibm.com>

* switch to code formula model v1.0.1 and new test pdf

Signed-off-by: Matteo-Omenetti <Matteo.Omenetti1@ibm.com>

* cleaned up the data folder in the tests

Signed-off-by: Peter Staar <taa@zurich.ibm.com>

* switch to code formula model v1.0.1 and new test pdf

Signed-off-by: Matteo-Omenetti <Matteo.Omenetti1@ibm.com>

* added three test-files for right-to-left

Signed-off-by: Peter Staar <taa@zurich.ibm.com>

* fix black

Signed-off-by: Matteo-Omenetti <Matteo.Omenetti1@ibm.com>

* added new gt for test_e2e_conversion

Signed-off-by: Matteo-Omenetti <Matteo.Omenetti1@ibm.com>

* added new gt for test_e2e_conversion

Signed-off-by: Matteo-Omenetti <Matteo.Omenetti1@ibm.com>

* Add code to expose text direction of cell

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

* new test file

Signed-off-by: Matteo-Omenetti <Matteo.Omenetti1@ibm.com>

* update lock

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* fix mypy reports

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* fix example filepaths

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* add test data results

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* pin wheel of latest docling-parse release

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* use latest docling-core

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* remove debugging code

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* fix path to files in example

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* Revert unwanted RTL additions

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

* Fix test data paths in examples

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

---------

Signed-off-by: Matteo-Omenetti <Matteo.Omenetti1@ibm.com>
Signed-off-by: Peter Staar <taa@zurich.ibm.com>
Signed-off-by: Christoph Auer <cau@zurich.ibm.com>
Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>
Co-authored-by: Matteo-Omenetti <Matteo.Omenetti1@ibm.com>
Co-authored-by: Peter Staar <taa@zurich.ibm.com>
Co-authored-by: Christoph Auer <cau@zurich.ibm.com>
This commit is contained in:
Michele Dolfi
2025-02-07 08:43:31 +01:00
committed by GitHub
parent ed74fe2ec0
commit 9114ada7bc
91 changed files with 620 additions and 313 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
docling/models/code_formula_model.py
View File

@@ -2,6 +2,7 @@ import re
from pathlib import Path
from typing import Iterable, List, Literal, Optional, Tuple, Union
import numpy as np
from docling_core.types.doc import (
CodeItem,
DocItemLabel,
@@ -103,7 +104,7 @@ class CodeFormulaModel(BaseItemAndImageEnrichmentModel):
artifacts_path = artifacts_path / self._model_repo_folder
self.code_formula_model = CodeFormulaPredictor(
artifacts_path=artifacts_path,
artifacts_path=str(artifacts_path),
device=device,
num_threads=accelerator_options.num_threads,
)
@@ -123,7 +124,7 @@ class CodeFormulaModel(BaseItemAndImageEnrichmentModel):
repo_id="ds4sd/CodeFormula",
force_download=force,
local_dir=local_dir,
revision="v1.0.0",
revision="v1.0.1",
)
return Path(download_path)
@@ -231,7 +232,7 @@ class CodeFormulaModel(BaseItemAndImageEnrichmentModel):
return
labels: List[str] = []
images: List[Image.Image] = []
images: List[Union[Image.Image, np.ndarray]] = []
elements: List[TextItem] = []
for el in element_batch:
assert isinstance(el.item, TextItem)

5
docling/models/document_picture_classifier.py
View File

@@ -1,6 +1,7 @@
from pathlib import Path
from typing import Iterable, List, Literal, Optional, Tuple, Union
import numpy as np
from docling_core.types.doc import (
DoclingDocument,
NodeItem,
@@ -94,7 +95,7 @@ class DocumentPictureClassifier(BaseEnrichmentModel):
artifacts_path = artifacts_path / self._model_repo_folder
self.document_picture_classifier = DocumentFigureClassifierPredictor(
artifacts_path=artifacts_path,
artifacts_path=str(artifacts_path),
device=device,
num_threads=accelerator_options.num_threads,
)
@@ -161,7 +162,7 @@ class DocumentPictureClassifier(BaseEnrichmentModel):
yield element
return
images: List[Image.Image] = []
images: List[Union[Image.Image, np.ndarray]] = []
elements: List[PictureItem] = []
for el in element_batch:
assert isinstance(el, PictureItem)

4
docling/models/layout_model.py
View File

@@ -150,10 +150,12 @@ class LayoutModel(BasePageModel):
else:
with TimeRecorder(conv_res, "layout"):
assert page.size is not None
page_image = page.get_image(scale=1.0)
assert page_image is not None
clusters = []
for ix, pred_item in enumerate(
self.layout_predictor.predict(page.get_image(scale=1.0))
self.layout_predictor.predict(page_image)
):
label = DocItemLabel(
pred_item["label"]