Merge pull request #1113 from myhloli/dev

refactor(ocr): remove unused functions and optimize OCR processing loop

magic_pdf/model/sub_modules/ocr/paddleocr/ocr_utils.py

@@ -1,5 +1,3 @@
-import math
-
 import numpy as np
 from loguru import logger
 
@@ -252,32 +250,6 @@ def get_ocr_result_list(ocr_res, useful_list):
     return ocr_result_list
 
 
-def calculate_angle_degrees(poly):
-    # 定义对角线的顶点
-    diagonal1 = (poly[0], poly[2])
-    diagonal2 = (poly[1], poly[3])
-
-    # 计算对角线的斜率
-    def slope(p1, p2):
-        return (p2[1] - p1[1]) / (p2[0] - p1[0]) if p2[0] != p1[0] else float('inf')
-
-    slope1 = slope(diagonal1[0], diagonal1[1])
-    slope2 = slope(diagonal2[0], diagonal2[1])
-
-    # 计算对角线与x轴的夹角（以弧度为单位）
-    angle1_radians = math.atan(slope1)
-    angle2_radians = math.atan(slope2)
-
-    # 将弧度转换为角度
-    angle1_degrees = math.degrees(angle1_radians)
-    angle2_degrees = math.degrees(angle2_radians)
-
-    # 取两条对角线与x轴夹角的平均值
-    average_angle_degrees = abs((angle1_degrees + angle2_degrees) / 2)
-    # logger.info(f"average_angle_degrees: {average_angle_degrees}")
-    return average_angle_degrees
-
-
 def calculate_is_angle(poly):
     p1, p2, p3, p4 = poly
     height = ((p4[1] - p1[1]) + (p3[1] - p2[1])) / 2

magic_pdf/model/sub_modules/ocr/paddleocr/ppocr_273_mod.py

@@ -63,7 +63,7 @@ class ModifiedPaddleOCR(PaddleOCR):
 
         if det and rec:
             ocr_res = []
-            for idx, img in enumerate(imgs):
+            for img in imgs:
                 img = preprocess_image(img)
                 dt_boxes, rec_res, _ = self.__call__(img, cls, mfd_res=mfd_res)
                 if not dt_boxes and not rec_res:
@@ -75,7 +75,7 @@ class ModifiedPaddleOCR(PaddleOCR):
             return ocr_res
         elif det and not rec:
             ocr_res = []
-            for idx, img in enumerate(imgs):
+            for img in imgs:
                 img = preprocess_image(img)
                 dt_boxes, elapse = self.text_detector(img)
                 if dt_boxes is None:
@@ -96,7 +96,7 @@ class ModifiedPaddleOCR(PaddleOCR):
         else:
             ocr_res = []
             cls_res = []
-            for idx, img in enumerate(imgs):
+            for img in imgs:
                 if not isinstance(img, list):
                     img = preprocess_image(img)
                     img = [img]

magic_pdf/pre_proc/ocr_detect_all_bboxes.py

@@ -1,11 +1,11 @@
-
 from magic_pdf.config.ocr_content_type import BlockType
 from magic_pdf.libs.boxbase import (
-    calculate_iou, calculate_overlap_area_in_bbox1_area_ratio,
+    calculate_iou,
+    calculate_overlap_area_in_bbox1_area_ratio,
     calculate_vertical_projection_overlap_ratio,
-    get_minbox_if_overlap_by_ratio)
-from magic_pdf.pre_proc.remove_bbox_overlap import \
-    remove_overlap_between_bbox_for_block
+    get_minbox_if_overlap_by_ratio
+)
+from magic_pdf.pre_proc.remove_bbox_overlap import remove_overlap_between_bbox_for_block
 
 
 def add_bboxes(blocks, block_type, bboxes):

magic_pdf/pre_proc/ocr_dict_merge.py

@@ -1,4 +1,3 @@
-
 from magic_pdf.config.ocr_content_type import BlockType, ContentType
 from magic_pdf.libs.boxbase import __is_overlaps_y_exceeds_threshold, calculate_overlap_area_in_bbox1_area_ratio
 
@@ -82,14 +81,6 @@ def fill_spans_in_blocks(blocks, spans, radio):
             if calculate_overlap_area_in_bbox1_area_ratio(
                     span_bbox, block_bbox) > radio:
                 block_spans.append(span)
-        '''行内公式调整, 高度调整至与同行文字高度一致(优先左侧, 其次右侧)'''
-        # displayed_list = []
-        # text_inline_lines = []
-        # modify_y_axis(block_spans, displayed_list, text_inline_lines)
-        '''模型识别错误的行间公式, type类型转换成行内公式'''
-        # block_spans = modify_inline_equation(block_spans, displayed_list, text_inline_lines)
-        '''bbox去除粘连'''  # 去粘连会影响span的bbox，导致后续fill的时候出错
-        # block_spans = remove_overlap_between_bbox_for_span(block_spans)
 
         block_dict['spans'] = block_spans
         block_with_spans.append(block_dict)
@@ -103,9 +94,6 @@ def fill_spans_in_blocks(blocks, spans, radio):
 
 
 def fix_block_spans_v2(block_with_spans):
-    """1、img_block和table_block因为包含caption和footnote的关系，存在block的嵌套关系
-    需要将caption和footnote的text_span放入相应img_block和table_block内的
-    caption_block和footnote_block中 2、同时需要删除block中的spans字段."""
     fix_blocks = []
     for block in block_with_spans:
         block_type = block['type']