Merge pull request #81 from myhloli/master

fix lost image or table bug

magic_pdf/dict2md/ocr_mkcontent.py

@@ -106,29 +106,30 @@ def ocr_mk_markdown_with_para_core_v2(paras_of_layout, mode, img_buket_path=""):
             if mode == 'nlp':
                 continue
             elif mode == 'mm':
-                for block in para_block['blocks']:
+                for block in para_block['blocks']:  # 1st.拼image_body
                     if block['type'] == BlockType.ImageBody:
                         for line in block['lines']:
                             for span in line['spans']:
                                 if span['type'] == ContentType.Image:
-                                    para_text = f"\n![]({join_path(img_buket_path, span['image_path'])})\n"
-                for block in para_block['blocks']:
+                                    para_text += f"\n![]({join_path(img_buket_path, span['image_path'])})\n"
+                for block in para_block['blocks']:  # 2nd.拼image_caption
                     if block['type'] == BlockType.ImageCaption:
                         para_text += merge_para_with_text(block)
         elif para_type == BlockType.Table:
             if mode == 'nlp':
                 continue
             elif mode == 'mm':
-                for block in para_block['blocks']:
+                for block in para_block['blocks']:  # 1st.拼table_caption
+                    if block['type'] == BlockType.TableCaption:
+                        para_text += merge_para_with_text(block)
+                for block in para_block['blocks']:  # 2nd.拼table_body
                     if block['type'] == BlockType.TableBody:
                         for line in block['lines']:
                             for span in line['spans']:
                                 if span['type'] == ContentType.Table:
-                                    para_text = f"\n![]({join_path(img_buket_path, span['image_path'])})\n"
-                for block in para_block['blocks']:
-                    if block['type'] == BlockType.TableCaption:
-                        para_text += merge_para_with_text(block)
-                    elif block['type'] == BlockType.TableFootnote:
+                                    para_text += f"\n![]({join_path(img_buket_path, span['image_path'])})\n"
+                for block in para_block['blocks']:  # 3rd.拼table_footnote
+                    if block['type'] == BlockType.TableFootnote:
                         para_text += merge_para_with_text(block)
 
         if para_text.strip() == '':

magic_pdf/pre_proc/ocr_dict_merge.py

@@ -156,7 +156,7 @@ def fill_spans_in_blocks(blocks, spans):
         block_spans = []
         for span in spans:
             span_bbox = span['bbox']
-            if calculate_overlap_area_in_bbox1_area_ratio(span_bbox, block_bbox) > 0.7:
+            if calculate_overlap_area_in_bbox1_area_ratio(span_bbox, block_bbox) > 0.6:
                 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)'''
@@ -167,8 +167,8 @@ def fill_spans_in_blocks(blocks, spans):
         '''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'''
-        block_spans = remove_overlap_between_bbox(block_spans)
+        '''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(block_spans)
 
         block_dict['spans'] = block_spans
         block_with_spans.append(block_dict)
@@ -208,7 +208,7 @@ def merge_spans_to_block(spans: list, block_bbox: list, block_type: str):
     block_spans = []
     # 如果有img_caption，则将img_block中的text_spans放入img_caption_block中
     for span in spans:
-        if calculate_overlap_area_in_bbox1_area_ratio(span['bbox'], block_bbox) > 0.8:
+        if calculate_overlap_area_in_bbox1_area_ratio(span['bbox'], block_bbox) > 0.6:
             block_spans.append(span)
     block_lines = merge_spans_to_line(block_spans)
     # 对line中的span进行排序
@@ -268,6 +268,7 @@ def fix_table_block(block, table_blocks):
     # 遍历table_blocks,找到与当前block匹配的table_block
     for table_block in table_blocks:
         if table_block['bbox'] == block['bbox']:
+
             # 创建table_body_block
             for span in block['spans']:
                 if span['type'] == ContentType.Table and span['bbox'] == table_block['table_body_bbox']:

magic_pdf/rw/S3ReaderWriter.py

 from magic_pdf.rw.AbsReaderWriter import AbsReaderWriter
-from magic_pdf.libs.commons import parse_aws_param, parse_bucket_key
+from magic_pdf.libs.commons import parse_aws_param, parse_bucket_key, join_path
 import boto3
 from loguru import logger
 from boto3.s3.transfer import TransferConfig
@@ -30,7 +30,7 @@ class S3ReaderWriter(AbsReaderWriter):
         if s3_relative_path.startswith("s3://"):
             s3_path = s3_relative_path
         else:
-            s3_path = os.path.join(self.path, s3_relative_path)
+            s3_path = join_path(self.path, s3_relative_path)
         bucket_name, key = parse_bucket_key(s3_path)
         res = self.client.get_object(Bucket=bucket_name, Key=key)
         body = res["Body"].read()
@@ -46,7 +46,7 @@ class S3ReaderWriter(AbsReaderWriter):
         if s3_relative_path.startswith("s3://"):
             s3_path = s3_relative_path
         else:
-            s3_path = os.path.join(self.path, s3_relative_path)
+            s3_path = join_path(self.path, s3_relative_path)
         if mode == MODE_TXT:
             body = content.encode(encoding)  # Encode text data as bytes
         elif mode == MODE_BIN:
@@ -61,7 +61,7 @@ class S3ReaderWriter(AbsReaderWriter):
         if path.startswith("s3://"):
             s3_path = path
         else:
-            s3_path = os.path.join(self.path, path)
+            s3_path = join_path(self.path, path)
         bucket_name, key = parse_bucket_key(s3_path)
 
         range_header = f'bytes={byte_start}-{byte_end}' if byte_end else f'bytes={byte_start}-'