对模型的layout坐标转换

magic_pdf/pdf_parse_by_ocr.py

@@ -25,15 +25,15 @@ def construct_page_component(page_id, blocks, layout_bboxes):
 
 
 def parse_pdf_by_ocr(
-    pdf_path,
-    s3_pdf_profile,
-    pdf_model_output,
-    book_name,
-    pdf_model_profile=None,
-    image_s3_config=None,
-    start_page_id=0,
-    end_page_id=None,
-    debug_mode=False,
+        pdf_path,
+        s3_pdf_profile,
+        pdf_model_output,
+        book_name,
+        pdf_model_profile=None,
+        image_s3_config=None,
+        start_page_id=0,
+        end_page_id=None,
+        debug_mode=False,
 ):
     pdf_bytes = read_file(pdf_path, s3_pdf_profile)
     save_tmp_path = os.path.join(os.path.dirname(__file__), "../..", "tmp", "unittest")
@@ -55,7 +55,6 @@ def parse_pdf_by_ocr(
         with open(pdf_local_path + ".pdf", "wb") as pdf_file:
             pdf_file.write(pdf_bytes)
 
-
     pdf_docs = fitz.open("pdf", pdf_bytes)
     # 初始化空的pdf_info_dict
     pdf_info_dict = {}
@@ -83,7 +82,8 @@ def parse_pdf_by_ocr(
         page_no_bboxes = parse_pageNos(page_id, page, ocr_page_info)
         header_bboxes = parse_headers(page_id, page, ocr_page_info)
         footer_bboxes = parse_footers(page_id, page, ocr_page_info)
-        footnote_bboxes = parse_footnotes_by_model(page_id, page, ocr_page_info, md_bookname_save_path, debug_mode=debug_mode)
+        footnote_bboxes = parse_footnotes_by_model(page_id, page, ocr_page_info, md_bookname_save_path,
+                                                   debug_mode=debug_mode)
 
         # 构建需要remove的bbox列表
         need_remove_spans_bboxes = []
@@ -103,7 +103,8 @@ def parse_pdf_by_ocr(
             allow_category_id_list = [1, 7, 13, 14, 15]
             if category_id in allow_category_id_list:
                 x0, y0, _, _, x1, y1, _, _ = layout_det['poly']
-                bbox = [int(x0/horizontal_scale_ratio), int(y0/vertical_scale_ratio), int(x1/horizontal_scale_ratio), int(y1/vertical_scale_ratio)]
+                bbox = [int(x0 / horizontal_scale_ratio), int(y0 / vertical_scale_ratio),
+                        int(x1 / horizontal_scale_ratio), int(y1 / vertical_scale_ratio)]
                 '''要删除的'''
                 #  3: 'header',      # 页眉
                 #  4: 'page number', # 页码
@@ -149,9 +150,11 @@ def parse_pdf_by_ocr(
 
         # 对tpye=["displayed_equation", "image", "table"]进行额外处理,如果左边有字的话,将该span的bbox中y0调整低于文字的y0
 
+        # 从ocr_page_info中解析layout信息(按自然阅读方向排序,并修复重叠和交错的bad case)
+        layout_bboxes = layout_detect(ocr_page_info['subfield_dets'], page)
 
-        # 将spans合并成line(从上到下,从左到右)
-        lines = merge_spans_to_line(spans)
+        # 将spans合并成line(在layout内,从上到下,从左到右)
+        lines = merge_spans_to_line(spans, layout_bboxes)
         # logger.info(lines)
 
         # 目前不做block拼接,先做个结构,每个block中只有一个line,block的bbox就是line的bbox
@@ -162,12 +165,8 @@ def parse_pdf_by_ocr(
                 "lines": [line],
             })
 
-        # 从ocr_page_info中解析layout信息(按自然阅读方向排序,并修复重叠和交错的bad case)
-        layout_bboxes = layout_detect(ocr_page_info['subfield_dets'])
-
         # 构造pdf_info_dict
         page_info = construct_page_component(page_id, blocks, layout_bboxes)
         pdf_info_dict[f"page_{page_id}"] = page_info
 
     return pdf_info_dict
-

magic_pdf/pre_proc/ocr_detect_layout.py

+import fitz
+
 from magic_pdf.libs.boxbase import _is_part_overlap, _is_in
+from magic_pdf.libs.coordinate_transform import get_scale_ratio
+
 
 def get_center_point(bbox):
     """
@@ -62,9 +66,7 @@ def adjust_layouts(layout_bboxes):
     return layout_bboxes
 
 
-
-
-def layout_detect(layout_info):
+def layout_detect(layout_info, page: fitz.Page):
     """
     对输入的布局信息进行解析，提取出每个子布局的边界框，并对所有子布局进行排序调整。
 
@@ -75,15 +77,18 @@ def layout_detect(layout_info):
         list: 经过排序调整后的所有子布局边界框信息的列表，每个边界框信息为字典类型，包含'layout_bbox'字段，表示边界框的坐标信息。
 
     """
+    horizontal_scale_ratio, vertical_scale_ratio = get_scale_ratio(layout_info, page)
     # 初始化布局边界框列表
     layout_bboxes = []
     # 遍历每个子布局
     for sub_layout in layout_info:
         # 提取子布局的边界框坐标信息
         x0, y0, _, _, x1, y1, _, _ = sub_layout['poly']
+        bbox = [int(x0 / horizontal_scale_ratio), int(y0 / vertical_scale_ratio),
+                int(x1 / horizontal_scale_ratio), int(y1 / vertical_scale_ratio)]
         # 创建子布局的边界框字典
         layout_bbox = {
-            "layout_bbox": [x0, y0, x1, y1],
+            "layout_bbox": bbox,
         }
         # 将子布局的边界框添加到列表中
         layout_bboxes.append(layout_bbox)
@@ -119,5 +124,3 @@ def layout_detect(layout_info):
 
     # 返回排序调整后的布局边界框列表
     return layout_bboxes
-
-