Skip to content

⚡️ Speed up function _merge_extracted_into_inferred_when_almost_the_same by 24% in PR #4112 (feat/track-text-source) #4114

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Nov 6, 2025
Merged

⚡️ Speed up function _merge_extracted_into_inferred_when_almost_the_same by 24% in PR #4112 (feat/track-text-source) #4114

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
24 changes: 16 additions & 8 deletions unstructured/partition/pdf_image/pdfminer_processing.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -57,14 +57,20 @@ def _validate_bbox(bbox: list[int | float]) -> bool:

def _minimum_containing_coords(*regions: TextRegions) -> np.ndarray:
# TODO: refactor to just use np array as input
return np.vstack(
# Optimization: Use np.stack and np.column_stack to build output in a single step
x1s = np.array([region.x1 for region in regions])
y1s = np.array([region.y1 for region in regions])
x2s = np.array([region.x2 for region in regions])
y2s = np.array([region.y2 for region in regions])
# Use np.min/max reduction rather than create matrix then operate. Transpose last for shape (N, 4)
return np.column_stack(
(
np.min([region.x1 for region in regions], axis=0),
np.min([region.y1 for region in regions], axis=0),
np.max([region.x2 for region in regions], axis=0),
np.max([region.y2 for region in regions], axis=0),
np.min(x1s, axis=0),
np.min(y1s, axis=0),
np.max(x2s, axis=0),
np.max(y2s, axis=0),
)
).T
)


def _inferred_is_elementtype(
Expand Down Expand Up @@ -120,7 +126,7 @@ def _merge_extracted_into_inferred_when_almost_the_same(
inferred_layout.element_coords,
threshold=same_region_threshold,
)
extracted_almost_the_same_as_inferred = boxes_almost_same.sum(axis=1).astype(bool)
extracted_almost_the_same_as_inferred = np.any(boxes_almost_same, axis=1)
# NOTE: if a row is full of False the argmax returns first index; we use the mask above to
# distinguish those (they would be False in the mask)
first_match = np.argmax(boxes_almost_same, axis=1)
Expand Down Expand Up @@ -584,7 +590,9 @@ def boxes_iou(
inter_area, boxa_area, boxb_area = areas_of_boxes_and_intersection_area(
coords1, coords2, round_to=round_to
)
return (inter_area / np.maximum(EPSILON_AREA, boxa_area + boxb_area.T - inter_area)) > threshold
denom = np.maximum(EPSILON_AREA, boxa_area + boxb_area.T - inter_area)
# Instead of (x/y) > t, use x > t*y for memory & speed with same result
return inter_area > (threshold * denom)


@requires_dependencies("unstructured_inference")
Expand Down