Always use pdfium coords

VikParuchuri · Oct 18, 2024 · 1263a1e · 1263a1e
1 parent 46ba8bf
commit 1263a1e
Showing 4 changed files with 10 additions and 26 deletions.
diff --git a/pdftext/inference.py b/pdftext/inference.py
@@ -45,7 +45,7 @@ def create_training_row(char_info, prev_char, currblock, currline):
 
     is_space = char in SPACES or char in TABS
 
-    return np.array([
+    return [
         char_center_x - currblock["center_x"],
         char_x1 - currblock_bbox[2],
         char_x1 - currblock_bbox[0],
@@ -65,9 +65,7 @@ def create_training_row(char_info, prev_char, currblock, currline):
         char_x2 - prev_x1,
         y_gap,
         char_y2 - prev_y1
-    ], dtype=np.float32)
-
-    return training_row
+    ]
 
 
 def update_span(line, span):
@@ -241,7 +239,7 @@ def inference(text_chars, model):
 
         training_idxs = sorted(training_data.keys())
         training_rows = [training_data[idx] for idx in training_idxs]
-        training_rows = np.stack(training_rows, axis=0)
+        training_rows = np.stack(training_rows, axis=0, dtype=np.float32)
 
         # Run inference
         predictions = model.run([output_name], {input_name: training_rows})[0]

diff --git a/pdftext/pdf/chars.py b/pdftext/pdf/chars.py
@@ -1,4 +1,5 @@
 import math
+from collections import defaultdict
 from typing import Dict, List
 
 import pypdfium2.raw as pdfium_c
@@ -41,7 +42,6 @@ def get_pdfium_chars(pdf, page_range, flatten_pdf, fontname_sample_freq=settings
             page = pdf.get_page(page_idx)
 
         text_page = page.get_textpage()
-        mediabox = page.get_mediabox()
         page_rotation = page.get_rotation()
         bbox = page.get_bbox()
         page_width = math.ceil(abs(bbox[2] - bbox[0]))
@@ -56,8 +56,6 @@ def get_pdfium_chars(pdf, page_range, flatten_pdf, fontname_sample_freq=settings
         if page_rotation == 90 or page_rotation == 270:
             page_width, page_height = page_height, page_width
 
-        bl_origin = (mediabox[0] == 0 and mediabox[1] == 0)
-
         text_chars = {
             "page": page_idx,
             "rotation": page_rotation,
@@ -87,7 +85,7 @@ def get_pdfium_chars(pdf, page_range, flatten_pdf, fontname_sample_freq=settings
             rotation = pdfium_c.FPDFText_GetCharAngle(text_page, i)
             rotation = rotation * rad_to_deg # convert from radians to degrees
             coords = text_page.get_charbox(i, loose=rotation == 0) # Loose doesn't work properly when charbox is rotated
-            device_coords = page_bbox_to_device_bbox(page, coords, page_width, page_height, bl_origin, page_rotation, normalize=True)
+            device_coords = page_bbox_to_device_bbox(page, coords, page_width, page_height, page_rotation, normalize=True)
 
             char_info = {
                 "font": {

diff --git a/pdftext/pdf/utils.py b/pdftext/pdf/utils.py
@@ -59,26 +59,14 @@ def pdfium_page_bbox_to_device_bbox(page, bbox, page_width, page_height, page_ro
     return dev_bbox
 
 
-def fast_page_bbox_to_device_bbox(page, bbox, page_width, page_height):
-    left, bottom, right, top = bbox
-
-    dev_bbox = [left, page_height-top, right, page_height-bottom]
-    return dev_bbox
-
-
-def page_bbox_to_device_bbox(page, bbox, page_width: int, page_height: int, bl_origin: bool, page_rotation: int, normalize=False):
+def page_bbox_to_device_bbox(page, bbox, page_width: int, page_height: int, page_rotation: int, normalize=False):
     orig_page_height, orig_page_width = page_height, page_width
     if page_rotation in [90, 270]:
         orig_page_height, orig_page_width = page_width, page_height
 
-    if bl_origin:
-        bbox = fast_page_bbox_to_device_bbox(page, bbox, page_width, page_height)
-        if page_rotation > 0:
-            bbox = rotate_page_bbox(bbox, page_rotation, page_width, page_height)
-    else:
-        bbox = pdfium_page_bbox_to_device_bbox(page, bbox, orig_page_width, orig_page_height, page_rotation)
-        if page_rotation > 0:
-            bbox = rotate_pdfium_bbox(bbox, page_rotation, page_width, page_height)
+    bbox = pdfium_page_bbox_to_device_bbox(page, bbox, orig_page_width, orig_page_height, page_rotation)
+    if page_rotation > 0:
+        bbox = rotate_pdfium_bbox(bbox, page_rotation, page_width, page_height)
 
     if normalize:
         bbox = [bbox[0] / page_width, bbox[1] / page_height, bbox[2] / page_width, bbox[3] / page_height]

diff --git a/pyproject.toml b/pyproject.toml
@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "pdftext"
-version = "0.3.16"
+version = "0.3.17"
 description = "Extract structured text from pdfs quickly"
 authors = ["Vik Paruchuri <vik.paruchuri@gmail.com>"]
 license = "Apache-2.0"