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

feat: Add pipeline timings and toggle visualization, establish debug settings (#183)

Browse Source 
* Add settings to turn visualization on or off

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

* Add profiling code to all models

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

* Refactor and fix profiling codes

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

* Visualization codes output PNG to debug dir

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

* Fixes for time logging

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

* Optimize imports

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

* Update lockfile

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

* Add start_timestamps to ProfilingItem

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
2024-10-30 15:04:19 +01:00
committed by GitHub
parent 94a5290789
commit 2a2c65bf4f
23 changed files with 998 additions and 771 deletions
Download Patch File Download Diff File Expand all files Collapse all files

84
docling/models/easyocr_model.py
View File

@@ -5,8 +5,11 @@ import numpy
from docling_core.types.doc import BoundingBox, CoordOrigin
from docling.datamodel.base_models import OcrCell, Page
from docling.datamodel.document import ConversionResult
from docling.datamodel.pipeline_options import EasyOcrOptions
from docling.datamodel.settings import settings
from docling.models.base_ocr_model import BaseOcrModel
from docling.utils.profiling import TimeRecorder
_log = logging.getLogger(__name__)
@@ -33,58 +36,65 @@ class EasyOcrModel(BaseOcrModel):
download_enabled=self.options.download_enabled,
)
def __call__(self, page_batch: Iterable[Page]) -> Iterable[Page]:
def __call__(
self, conv_res: ConversionResult, page_batch: Iterable[Page]
) -> Iterable[Page]:
if not self.enabled:
yield from page_batch
return
for page in page_batch:
assert page._backend is not None
if not page._backend.is_valid():
yield page
else:
ocr_rects = self.get_ocr_rects(page)
with TimeRecorder(conv_res, "ocr"):
ocr_rects = self.get_ocr_rects(page)
all_ocr_cells = []
for ocr_rect in ocr_rects:
# Skip zero area boxes
if ocr_rect.area() == 0:
continue
high_res_image = page._backend.get_page_image(
scale=self.scale, cropbox=ocr_rect
)
im = numpy.array(high_res_image)
result = self.reader.readtext(im)
del high_res_image
del im
cells = [
OcrCell(
id=ix,
text=line[1],
confidence=line[2],
bbox=BoundingBox.from_tuple(
coord=(
(line[0][0][0] / self.scale) + ocr_rect.l,
(line[0][0][1] / self.scale) + ocr_rect.t,
(line[0][2][0] / self.scale) + ocr_rect.l,
(line[0][2][1] / self.scale) + ocr_rect.t,
),
origin=CoordOrigin.TOPLEFT,
),
all_ocr_cells = []
for ocr_rect in ocr_rects:
# Skip zero area boxes
if ocr_rect.area() == 0:
continue
high_res_image = page._backend.get_page_image(
scale=self.scale, cropbox=ocr_rect
)
for ix, line in enumerate(result)
]
all_ocr_cells.extend(cells)
im = numpy.array(high_res_image)
result = self.reader.readtext(im)
## Remove OCR cells which overlap with programmatic cells.
filtered_ocr_cells = self.filter_ocr_cells(all_ocr_cells, page.cells)
del high_res_image
del im
page.cells.extend(filtered_ocr_cells)
cells = [
OcrCell(
id=ix,
text=line[1],
confidence=line[2],
bbox=BoundingBox.from_tuple(
coord=(
(line[0][0][0] / self.scale) + ocr_rect.l,
(line[0][0][1] / self.scale) + ocr_rect.t,
(line[0][2][0] / self.scale) + ocr_rect.l,
(line[0][2][1] / self.scale) + ocr_rect.t,
),
origin=CoordOrigin.TOPLEFT,
),
)
for ix, line in enumerate(result)
]
all_ocr_cells.extend(cells)
## Remove OCR cells which overlap with programmatic cells.
filtered_ocr_cells = self.filter_ocr_cells(
all_ocr_cells, page.cells
)
page.cells.extend(filtered_ocr_cells)
# DEBUG code:
# self.draw_ocr_rects_and_cells(page, ocr_rects)
if settings.debug.visualize_ocr:
self.draw_ocr_rects_and_cells(conv_res, page, ocr_rects)
yield page