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

104
docling/models/tesseract_ocr_model.py
View File

@@ -4,8 +4,11 @@ from typing import Iterable
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 TesseractOcrOptions
from docling.datamodel.settings import settings
from docling.models.base_ocr_model import BaseOcrModel
from docling.utils.profiling import TimeRecorder
_log = logging.getLogger(__name__)
@@ -61,7 +64,9 @@ class TesseractOcrModel(BaseOcrModel):
# Finalize the tesseractAPI
self.reader.End()
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
@@ -72,59 +77,66 @@ class TesseractOcrModel(BaseOcrModel):
if not page._backend.is_valid():
yield page
else:
assert self.reader is not None
with TimeRecorder(conv_res, "ocr"):
ocr_rects = self.get_ocr_rects(page)
assert self.reader is not None
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
)
ocr_rects = self.get_ocr_rects(page)
# Retrieve text snippets with their bounding boxes
self.reader.SetImage(high_res_image)
boxes = self.reader.GetComponentImages(
self.reader_RIL.TEXTLINE, True
)
cells = []
for ix, (im, box, _, _) in enumerate(boxes):
# Set the area of interest. Tesseract uses Bottom-Left for the origin
self.reader.SetRectangle(box["x"], box["y"], box["w"], box["h"])
# Extract text within the bounding box
text = self.reader.GetUTF8Text().strip()
confidence = self.reader.MeanTextConf()
left = box["x"] / self.scale
bottom = box["y"] / self.scale
right = (box["x"] + box["w"]) / self.scale
top = (box["y"] + box["h"]) / self.scale
cells.append(
OcrCell(
id=ix,
text=text,
confidence=confidence,
bbox=BoundingBox.from_tuple(
coord=(left, top, right, bottom),
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
)
# del high_res_image
all_ocr_cells.extend(cells)
# Retrieve text snippets with their bounding boxes
self.reader.SetImage(high_res_image)
boxes = self.reader.GetComponentImages(
self.reader_RIL.TEXTLINE, True
)
## Remove OCR cells which overlap with programmatic cells.
filtered_ocr_cells = self.filter_ocr_cells(all_ocr_cells, page.cells)
cells = []
for ix, (im, box, _, _) in enumerate(boxes):
# Set the area of interest. Tesseract uses Bottom-Left for the origin
self.reader.SetRectangle(
box["x"], box["y"], box["w"], box["h"]
)
page.cells.extend(filtered_ocr_cells)
# Extract text within the bounding box
text = self.reader.GetUTF8Text().strip()
confidence = self.reader.MeanTextConf()
left = box["x"] / self.scale
bottom = box["y"] / self.scale
right = (box["x"] + box["w"]) / self.scale
top = (box["y"] + box["h"]) / self.scale
cells.append(
OcrCell(
id=ix,
text=text,
confidence=confidence,
bbox=BoundingBox.from_tuple(
coord=(left, top, right, bottom),
origin=CoordOrigin.TOPLEFT,
),
)
)
# del high_res_image
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