Merge pull request #4677 from opendatalab/master

master->dev

docs/en/quick_start/index.md

@@ -50,7 +50,7 @@ A WebUI developed based on Gradio, with a simple interface and only core parsing
     </tr> 
     <tr>
       <th>Accuracy<sup>1</sup></th>
-      <td style="text-align:center;">82+</td>
+      <td style="text-align:center;">86+</td>
       <td colspan="4" style="text-align:center;">90+</td>
     </tr>
     <tr>
@@ -70,15 +70,15 @@ A WebUI developed based on Gradio, with a simple interface and only core parsing
     </tr>
     <tr>
       <th>Min VRAM</th>
-      <td style="text-align:center;">6GB</td>
-      <td style="text-align:center;">10GB</td>
+      <td style="text-align:center;">4GB</td>
       <td style="text-align:center;">8GB</td>
-      <td style="text-align:center;">3GB</td>
+      <td style="text-align:center;">8GB</td>
+      <td style="text-align:center;">2GB</td>
     </tr>
     <tr>
       <th>RAM</th>
       <td colspan="3" style="text-align:center;">Min 16GB+, Recommended 32GB+</td>
-      <td colspan="2" style="text-align:center;">8GB</td>
+      <td colspan="2" style="text-align:center;">16GB</td>
     </tr>
     <tr>
       <th>Disk Space</th>

docs/zh/quick_start/index.md

@@ -50,7 +50,7 @@
     </tr> 
     <tr>
       <th>精度指标<sup>1</sup></th>
-      <td style="text-align:center;">82+</td>
+      <td style="text-align:center;">86+</td>
       <td colspan="4" style="text-align:center;">90+</td>
     </tr>
     <tr>
@@ -70,15 +70,15 @@
     </tr>
     <tr>
       <th>显存最低要求</th>
-      <td style="text-align:center;">6GB</td>
-      <td style="text-align:center;">10GB</td>
+      <td style="text-align:center;">4GB</td>
       <td style="text-align:center;">8GB</td>
-      <td style="text-align:center;">3GB</td>
+      <td style="text-align:center;">8GB</td>
+      <td style="text-align:center;">2GB</td>
     </tr>
     <tr>
       <th>内存要求</th>
       <td colspan="3" style="text-align:center;">最低16GB以上,推荐32GB以上</td>
-      <td colspan="2" style="text-align:center;">8GB</td>
+      <td colspan="2" style="text-align:center;">16GB</td>
     </tr>
     <tr>
       <th>磁盘空间要求</th>

mineru/backend/hybrid/hybrid_model_output_to_middle_json.py

@@ -131,23 +131,27 @@ def blocks_to_page_info(
     return page_info
 
 
+def _iter_block_spans(block):
+    for line in block.get("lines", []):
+        for span in line.get("spans", []):
+            yield span
+
+    for sub_block in block.get("blocks", []):
+        yield from _iter_block_spans(sub_block)
+
+
 def _apply_post_ocr(pdf_info_list, hybrid_pipeline_model):
     need_ocr_list = []
     img_crop_list = []
-    text_block_list = []
     for page_info in pdf_info_list:
-        for block in page_info['para_blocks']:
-            if block['type'] in ['table', 'image', 'list', 'code']:
-                for sub_block in block['blocks']:
-                    if not sub_block['type'].endswith('body'):
-                        text_block_list.append(sub_block)
-            elif block['type'] in ['text', 'title', 'ref_text']:
-                text_block_list.append(block)
-        for block in page_info['discarded_blocks']:
-            text_block_list.append(block)
-    for block in text_block_list:
-        for line in block['lines']:
-            for span in line['spans']:
+        for block in page_info.get('para_blocks', []):
+            for span in _iter_block_spans(block):
+                if 'np_img' in span:
+                    need_ocr_list.append(span)
+                    img_crop_list.append(rotate_vertical_crop_if_needed(span['np_img']))
+                    span.pop('np_img')
+        for block in page_info.get('discarded_blocks', []):
+            for span in _iter_block_spans(block):
                 if 'np_img' in span:
                     need_ocr_list.append(span)
                     img_crop_list.append(rotate_vertical_crop_if_needed(span['np_img']))

mineru/backend/pipeline/model_json_to_middle_json.py

@@ -203,6 +203,15 @@ def _extract_text_from_block(block):
     return "".join(text_parts).strip()
 
 
+def _iter_block_spans(block):
+    for line in block.get("lines", []):
+        for span in line.get("spans", []):
+            yield span
+
+    for sub_block in block.get("blocks", []):
+        yield from _iter_block_spans(sub_block)
+
+
 def _normalize_formula_tag_content(tag_content):
     tag_content = full_to_half(tag_content.strip())
     if tag_content.startswith("("):
@@ -261,22 +270,18 @@ def _optimize_formula_number_blocks(pdf_info_list):
 def _apply_post_ocr(pdf_info_list, lang=None):
     need_ocr_list = []
     img_crop_list = []
-    text_block_list = []
 
     for page_info in pdf_info_list:
-        for block in page_info['preproc_blocks']:
-            if 'blocks' in block:
-                for sub_block in block['blocks']:
-                    if sub_block.get("type", "").endswith('caption') or sub_block.get("type", "").endswith('footnote'):
-                        text_block_list.append(sub_block)
-            elif block["type"] not in [BlockType.INTERLINE_EQUATION, BlockType.SEAL]:
-                text_block_list.append(block)
-        for block in page_info['discarded_blocks']:
-            text_block_list.append(block)
+        for block in page_info.get('preproc_blocks', []):
+            for span in _iter_block_spans(block):
+                if 'np_img' in span:
+                    need_ocr_list.append(span)
+                    # Keep post-OCR rec aligned with the main OCR pipeline for vertical tall crops.
+                    img_crop_list.append(rotate_vertical_crop_if_needed(span['np_img']))
+                    span.pop('np_img')
 
-    for block in text_block_list:
-        for line in block['lines']:
-            for span in line['spans']:
+        for block in page_info.get('discarded_blocks', []):
+            for span in _iter_block_spans(block):
                 if 'np_img' in span:
                     need_ocr_list.append(span)
                     # Keep post-OCR rec aligned with the main OCR pipeline for vertical tall crops.

mineru/version.py

@@ -1 +1 @@
-__version__ = "2.7.6"
+__version__ = "3.0.1"