Merge pull request #969 from opendatalab/release-0.9.3

Release 0.9.3

magic_pdf/model/ppTableModel.py → magic_pdf/model/sub_modules/table/tablemaster/tablemaster_paddle.py

@@ -7,7 +7,7 @@ from PIL import Image
 import numpy as np
 
 
-class ppTableModel(object):
+class TableMasterPaddleModel(object):
     """
         This class is responsible for converting image of table into HTML format using a pre-trained model.
 

magic_pdf/para/para_split_v3.py

@@ -77,14 +77,12 @@ def __is_list_or_index_block(block):
 
         # 如果首行左边不顶格而右边顶格,末行左边顶格而右边不顶格 （第一行可能可以右边不顶格）
         if (first_line['bbox'][0] - block['bbox_fs'][0] > line_height / 2 and
-                # block['bbox_fs'][2] - first_line['bbox'][2] < line_height and
                 abs(last_line['bbox'][0] - block['bbox_fs'][0]) < line_height / 2 and
                 block['bbox_fs'][2] - last_line['bbox'][2] > line_height
         ):
             multiple_para_flag = True
 
         for line in block['lines']:
-
             line_mid_x = (line['bbox'][0] + line['bbox'][2]) / 2
             block_mid_x = (block['bbox_fs'][0] + block['bbox_fs'][2]) / 2
             if (
@@ -102,13 +100,13 @@ def __is_list_or_index_block(block):
                 if span_type == ContentType.Text:
                     line_text += span['content'].strip()
 
+            # 添加所有文本，包括空行，保持与block['lines']长度一致
             lines_text_list.append(line_text)
 
             # 计算line左侧顶格数量是否大于2，是否顶格用abs(block['bbox_fs'][0] - line['bbox'][0]) < line_height/2 来判断
             if abs(block['bbox_fs'][0] - line['bbox'][0]) < line_height / 2:
                 left_close_num += 1
             elif line['bbox'][0] - block['bbox_fs'][0] > line_height:
-                # logger.info(f"{line_text}, {block['bbox_fs']}, {line['bbox']}")
                 left_not_close_num += 1
 
             # 计算右侧是否顶格
@@ -117,7 +115,6 @@ def __is_list_or_index_block(block):
             else:
                 # 右侧不顶格情况下是否有一段距离，拍脑袋用0.3block宽度做阈值
                 closed_area = 0.26 * block_weight
-                # closed_area = 5 * line_height
                 if block['bbox_fs'][2] - line['bbox'][2] > closed_area:
                     right_not_close_num += 1
 
@@ -128,6 +125,7 @@ def __is_list_or_index_block(block):
         num_start_count = 0
         num_end_count = 0
         flag_end_count = 0
+
         if len(lines_text_list) > 0:
             for line_text in lines_text_list:
                 if len(line_text) > 0:
@@ -138,11 +136,10 @@ def __is_list_or_index_block(block):
                     if line_text[-1].isdigit():
                         num_end_count += 1
 
-            if flag_end_count / len(lines_text_list) >= 0.8:
-                line_end_flag = True
-
             if num_start_count / len(lines_text_list) >= 0.8 or num_end_count / len(lines_text_list) >= 0.8:
                 line_num_flag = True
+            if flag_end_count / len(lines_text_list) >= 0.8:
+                line_end_flag = True
 
         # 有的目录右侧不贴边, 目前认为左边或者右边有一边全贴边，且符合数字规则极为index
         if ((left_close_num / len(block['lines']) >= 0.8 or right_close_num / len(block['lines']) >= 0.8)
@@ -176,7 +173,7 @@ def __is_list_or_index_block(block):
                 # 这种是大部分line item 都有结束标识符的情况，按结束标识符区分不同item
                 elif line_end_flag:
                     for i, line in enumerate(block['lines']):
-                        if lines_text_list[i][-1] in LIST_END_FLAG:
+                        if len(lines_text_list[i]) > 0 and lines_text_list[i][-1] in LIST_END_FLAG:
                             line[ListLineTag.IS_LIST_END_LINE] = True
                             if i + 1 < len(block['lines']):
                                 block['lines'][i + 1][ListLineTag.IS_LIST_START_LINE] = True
@@ -187,13 +184,14 @@ def __is_list_or_index_block(block):
                         if line_start_flag:
                             line[ListLineTag.IS_LIST_START_LINE] = True
                             line_start_flag = False
-                        # elif abs(block['bbox_fs'][2] - line['bbox'][2]) > line_height:
+
                         if abs(block['bbox_fs'][2] - line['bbox'][2]) > 0.1 * block_weight:
                             line[ListLineTag.IS_LIST_END_LINE] = True
                             line_start_flag = True
-            # 一种有缩进的特殊有序list,start line 左侧不贴边且以数字开头，end line 以 IS_LIST_END_LINE 结尾且数量和start line 一致
-            elif num_start_count >= 2 and num_start_count == flag_end_count:  # 简单一点先不考虑左侧不贴边的情况
+            # 一种有缩进的特殊有序list,start line 左侧不贴边且以数字开头，end line 以 IS_LIST_END_FLAG 结尾且数量和start line 一致
+            elif num_start_count >= 2 and num_start_count == flag_end_count:
                 for i, line in enumerate(block['lines']):
+                    if len(lines_text_list[i]) > 0:
                         if lines_text_list[i][0].isdigit():
                             line[ListLineTag.IS_LIST_START_LINE] = True
                         if lines_text_list[i][-1] in LIST_END_FLAG:

magic_pdf/pdf_parse_union_core_v2.py

@@ -30,8 +30,8 @@ from magic_pdf.pre_proc.equations_replace import (
 from magic_pdf.pre_proc.ocr_detect_all_bboxes import \
     ocr_prepare_bboxes_for_layout_split_v2
 from magic_pdf.pre_proc.ocr_dict_merge import (fill_spans_in_blocks,
-                                               fix_block_spans,
-                                               fix_discarded_block, fix_block_spans_v2)
+                                               fix_discarded_block,
+                                               fix_block_spans_v2)
 from magic_pdf.pre_proc.ocr_span_list_modify import (
     get_qa_need_list_v2, remove_overlaps_low_confidence_spans,
     remove_overlaps_min_spans)
@@ -164,7 +164,7 @@ class ModelSingleton:
 
 
 def do_predict(boxes: List[List[int]], model) -> List[int]:
-    from magic_pdf.model.v3.helpers import (boxes2inputs, parse_logits,
+    from magic_pdf.model.sub_modules.reading_oreder.layoutreader.helpers import (boxes2inputs, parse_logits,
                                                                                  prepare_inputs)
 
     inputs = boxes2inputs(boxes)
@@ -174,8 +174,10 @@ def do_predict(boxes: List[List[int]], model) -> List[int]:
 
 
 def cal_block_index(fix_blocks, sorted_bboxes):
-    for block in fix_blocks:
 
+    if sorted_bboxes is not None:
+        # 使用layoutreader排序
+        for block in fix_blocks:
             line_index_list = []
             if len(block['lines']) == 0:
                 block['index'] = sorted_bboxes.index(block['bbox'])
@@ -191,6 +193,38 @@ def cal_block_index(fix_blocks, sorted_bboxes):
                 block['virtual_lines'] = copy.deepcopy(block['lines'])
                 block['lines'] = copy.deepcopy(block['real_lines'])
                 del block['real_lines']
+    else:
+        # 使用xycut排序
+        block_bboxes = []
+        for block in fix_blocks:
+            block_bboxes.append(block['bbox'])
+
+            # 删除图表body block中的虚拟line信息, 并用real_lines信息回填
+            if block['type'] in [BlockType.ImageBody, BlockType.TableBody]:
+                block['virtual_lines'] = copy.deepcopy(block['lines'])
+                block['lines'] = copy.deepcopy(block['real_lines'])
+                del block['real_lines']
+
+        import numpy as np
+        from magic_pdf.model.sub_modules.reading_oreder.layoutreader.xycut import recursive_xy_cut
+
+        random_boxes = np.array(block_bboxes)
+        np.random.shuffle(random_boxes)
+        res = []
+        recursive_xy_cut(np.asarray(random_boxes).astype(int), np.arange(len(block_bboxes)), res)
+        assert len(res) == len(block_bboxes)
+        sorted_boxes = random_boxes[np.array(res)].tolist()
+
+        for i, block in enumerate(fix_blocks):
+            block['index'] = sorted_boxes.index(block['bbox'])
+
+        # 生成line index
+        sorted_blocks = sorted(fix_blocks, key=lambda b: b['index'])
+        line_inedx = 1
+        for block in sorted_blocks:
+            for line in block['lines']:
+                line['index'] = line_inedx
+                line_inedx += 1
 
     return fix_blocks
 
@@ -264,6 +298,9 @@ def sort_lines_by_model(fix_blocks, page_w, page_h, line_height):
                 block['lines'].append({'bbox': line, 'spans': []})
             page_line_list.extend(lines)
 
+    if len(page_line_list) > 200:  # layoutreader最高支持512line
+        return None
+
     # 使用layoutreader排序
     x_scale = 1000.0 / page_w
     y_scale = 1000.0 / page_h

magic_pdf/resources/model_config/model_configs.yaml

@@ -5,3 +5,4 @@ weights:
   unimernet_small: MFR/unimernet_small
   struct_eqtable: TabRec/StructEqTable
   tablemaster: TabRec/TableMaster
+  rapid_table: TabRec/RapidTable
\ No newline at end of file

magic_pdf/tools/common.py

@@ -14,6 +14,9 @@ from magic_pdf.pipe.TXTPipe import TXTPipe
 from magic_pdf.pipe.UNIPipe import UNIPipe
 from magic_pdf.rw.AbsReaderWriter import AbsReaderWriter
 from magic_pdf.rw.DiskReaderWriter import DiskReaderWriter
+import fitz
+# from io import BytesIO
+# from pypdf import PdfReader, PdfWriter
 
 
 def prepare_env(output_dir, pdf_file_name, method):
@@ -26,6 +29,42 @@ def prepare_env(output_dir, pdf_file_name, method):
     return local_image_dir, local_md_dir
 
 
+# def convert_pdf_bytes_to_bytes_by_pypdf(pdf_bytes, start_page_id=0, end_page_id=None):
+#     # 将字节数据包装在 BytesIO 对象中
+#     pdf_file = BytesIO(pdf_bytes)
+#     # 读取 PDF 的字节数据
+#     reader = PdfReader(pdf_file)
+#     # 创建一个新的 PDF 写入器
+#     writer = PdfWriter()
+#     # 将所有页面添加到新的 PDF 写入器中
+#     end_page_id = end_page_id if end_page_id is not None and end_page_id >= 0 else len(reader.pages) - 1
+#     if end_page_id > len(reader.pages) - 1:
+#         logger.warning("end_page_id is out of range, use pdf_docs length")
+#         end_page_id = len(reader.pages) - 1
+#     for i, page in enumerate(reader.pages):
+#         if start_page_id <= i <= end_page_id:
+#             writer.add_page(page)
+#     # 创建一个字节缓冲区来存储输出的 PDF 数据
+#     output_buffer = BytesIO()
+#     # 将 PDF 写入字节缓冲区
+#     writer.write(output_buffer)
+#     # 获取字节缓冲区的内容
+#     converted_pdf_bytes = output_buffer.getvalue()
+#     return converted_pdf_bytes
+
+
+def convert_pdf_bytes_to_bytes_by_pymupdf(pdf_bytes, start_page_id=0, end_page_id=None):
+    document = fitz.open("pdf", pdf_bytes)
+    output_document = fitz.open()
+    end_page_id = end_page_id if end_page_id is not None and end_page_id >= 0 else len(document) - 1
+    if end_page_id > len(document) - 1:
+        logger.warning("end_page_id is out of range, use pdf_docs length")
+        end_page_id = len(document) - 1
+    output_document.insert_pdf(document, from_page=start_page_id, to_page=end_page_id)
+    output_bytes = output_document.tobytes()
+    return output_bytes
+
+
 def do_parse(
     output_dir,
     pdf_file_name,
@@ -55,6 +94,8 @@ def do_parse(
         f_draw_model_bbox = True
         f_draw_line_sort_bbox = True
 
+    pdf_bytes = convert_pdf_bytes_to_bytes_by_pymupdf(pdf_bytes, start_page_id, end_page_id)
+
     orig_model_list = copy.deepcopy(model_list)
     local_image_dir, local_md_dir = prepare_env(output_dir, pdf_file_name,
                                                 parse_method)
@@ -66,15 +107,18 @@ def do_parse(
     if parse_method == 'auto':
         jso_useful_key = {'_pdf_type': '', 'model_list': model_list}
         pipe = UNIPipe(pdf_bytes, jso_useful_key, image_writer, is_debug=True,
-                       start_page_id=start_page_id, end_page_id=end_page_id, lang=lang,
+                       # start_page_id=start_page_id, end_page_id=end_page_id,
+                       lang=lang,
                        layout_model=layout_model, formula_enable=formula_enable, table_enable=table_enable)
     elif parse_method == 'txt':
         pipe = TXTPipe(pdf_bytes, model_list, image_writer, is_debug=True,
-                       start_page_id=start_page_id, end_page_id=end_page_id, lang=lang,
+                       # start_page_id=start_page_id, end_page_id=end_page_id,
+                       lang=lang,
                        layout_model=layout_model, formula_enable=formula_enable, table_enable=table_enable)
     elif parse_method == 'ocr':
         pipe = OCRPipe(pdf_bytes, model_list, image_writer, is_debug=True,
-                       start_page_id=start_page_id, end_page_id=end_page_id, lang=lang,
+                       # start_page_id=start_page_id, end_page_id=end_page_id,
+                       lang=lang,
                        layout_model=layout_model, formula_enable=formula_enable, table_enable=table_enable)
     else:
         logger.error('unknown parse method')

next_docs/en/_static/image/ReadTheDocs.svg

+<?xml version="1.0" encoding="UTF-8" standalone="no" ?>
+<!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN" "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd">
+<svg xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" version="1.1" width="224" height="72" viewBox="-29 -3.67 224 72" xml:space="preserve">
+<desc>Created with Fabric.js 5.2.4</desc>
+<defs>
+</defs>
+<rect x="0" y="0" width="100%" height="100%" fill="transparent"></rect>
+<g transform="matrix(1 0 0 1 112 36)" id="7a867f58-a908-4f30-a839-fb725512b521"  >
+<rect style="stroke: none; stroke-width: 1; stroke-dasharray: none; stroke-linecap: butt; stroke-dashoffset: 0; stroke-linejoin: miter; stroke-miterlimit: 4; fill: rgb(255,255,255); fill-rule: nonzero; opacity: 1; visibility: hidden;" vector-effect="non-scaling-stroke"  x="-112" y="-36" rx="0" ry="0" width="224" height="72" />
+</g>
+<g transform="matrix(Infinity NaN NaN Infinity 0 0)" id="29611287-bf1c-4faf-8eb1-df32f6424829"  >
+</g>
+<g transform="matrix(0.07 0 0 0.07 382.02 122.8)" id="60cdd44f-027a-437a-92c4-c8d44c60ef9e"  >
+<path style="stroke: rgb(0,0,0); stroke-width: 0; stroke-dasharray: none; stroke-linecap: butt; stroke-dashoffset: 0; stroke-linejoin: miter; stroke-miterlimit: 4; fill: rgb(50,50,42); fill-rule: nonzero; opacity: 1;" vector-effect="non-scaling-stroke"  transform=" translate(-64, -64)" d="M 57.62 61.68 C 55.919999999999995 61.92 54.75 63.46 55 65.11 C 55.1668510745875 66.32380621250819 56.039448735907676 67.32218371690155 57.22 67.65 C 57.22 67.65 64.69 70.11 77.4 71.16000000000001 C 87.61000000000001 72.01 99.2 70.43 99.2 70.43 C 100.9 70.39 102.23 68.98 102.19 67.28 C 102.17037752125772 66.4652516996782 101.82707564255573 65.69186585376654 101.23597809465886 65.13079230830253 C 100.644880546762 64.56971876283853 99.85466451370849 64.26716220997277 99.03999999999999 64.29 C 98.83999999999999 64.29 98.63999999999999 64.33000000000001 98.42999999999999 64.37 C 98.42999999999999 64.37 87.08999999999999 65.78 77.88 65.02000000000001 C 65.72999999999999 64.05000000000001 59.11 61.83000000000001 59.11 61.83000000000001 C 58.63 61.670000000000016 58.1 61.59000000000001 57.62 61.670000000000016 Z M 57.62 46.46 C 55.919999999999995 46.7 54.75 48.24 55 49.89 C 55.1668510745875 51.10380621250818 56.039448735907676 52.10218371690154 57.22 52.43 C 57.22 52.43 64.69 54.89 77.4 55.94 C 87.61000000000001 56.79 99.2 55.21 99.2 55.21 C 100.9 55.17 102.23 53.76 102.19 52.06 C 102.17037752125772 51.245251699678214 101.82707564255573 50.47186585376654 101.23597809465886 49.91079230830253 C 100.644880546762 49.34971876283853 99.85466451370849 49.047162209972754 99.03999999999999 49.07 C 98.83999999999999 49.07 98.63999999999999 49.11 98.42999999999999 49.15 C 98.42999999999999 49.15 87.08999999999999 50.559999999999995 77.88 49.8 C 65.72999999999999 48.83 59.11 46.61 59.11 46.61 C 58.63 46.45 58.1 46.37 57.62 46.45 Z M 57.62 31.240000000000002 C 55.919999999999995 31.48 54.75 33.02 55 34.67 C 55.1668510745875 35.88380621250818 56.039448735907676 36.882183716901544 57.22 37.21 C 57.22 37.21 64.69 39.67 77.4 40.72 C 87.61000000000001 41.57 99.2 39.99 99.2 39.99 C 100.9 39.95 102.23 38.54 102.19 36.84 C 102.17037752125772 36.025251699678215 101.82707564255573 35.25186585376654 101.23597809465886 34.690792308302534 C 100.644880546762 34.12971876283853 99.85466451370849 33.827162209972755 99.03999999999999 33.85 C 98.83999999999999 33.85 98.63999999999999 33.89 98.42999999999999 33.93 C 98.42999999999999 33.93 87.08999999999999 35.339999999999996 77.88 34.58 C 65.72999999999999 33.61 59.11 31.389999999999997 59.11 31.389999999999997 C 58.63 31.229999999999997 58.1 31.189999999999998 57.62 31.229999999999997 Z M 57.62 16.060000000000002 C 55.919999999999995 16.3 54.75 17.840000000000003 55 19.490000000000002 C 55.1668510745875 20.703806212508187 56.039448735907676 21.702183716901544 57.22 22.03 C 57.22 22.03 64.69 24.490000000000002 77.4 25.54 C 87.61000000000001 26.39 99.2 24.81 99.2 24.81 C 100.9 24.77 102.23 23.36 102.19 21.66 C 102.17037752125772 20.84525169967821 101.82707564255573 20.07186585376654 101.23597809465886 19.510792308302534 C 100.644880546762 18.949718762838526 99.8546645137085 18.64716220997276 99.03999999999999 18.67 C 98.83999999999999 18.67 98.63999999999999 18.71 98.42999999999999 18.75 C 98.42999999999999 18.75 87.08999999999999 20.16 77.88 19.4 C 65.72999999999999 18.43 59.11 16.209999999999997 59.11 16.209999999999997 C 58.637850878541954 16.01924514007714 58.12188500879498 15.963839409097599 57.62 16.049999999999997 Z M 36.31 0 C 20.32 0.12 14.39 5.05 14.39 5.05 L 14.39 124.42 C 14.39 124.42 20.2 119.41 38.93 120.18 C 57.66 120.95000000000002 61.5 127.53 84.50999999999999 127.97000000000001 C 107.52 128.41000000000003 113.28999999999999 124.42000000000002 113.28999999999999 124.42000000000002 L 113.60999999999999 2.750000000000014 C 113.60999999999999 2.750000000000014 103.28 5.7 83.09 5.86 C 62.95 6.01 58.11 0.73 39.62 0.12 C 38.49 0.04 37.4 0 36.31 0 Z M 49.67 7.79 C 49.67 7.79 59.36 10.98 77.24000000000001 11.870000000000001 C 92.38000000000001 12.64 107.52000000000001 10.38 107.52000000000001 10.38 L 107.52000000000001 118.53 C 107.52000000000001 118.53 99.85000000000001 122.57000000000001 80.68 121.19 C 65.82000000000001 120.14 49.480000000000004 114.49 49.480000000000004 114.49 L 49.68000000000001 7.799999999999997 Z M 40.35 10.620000000000001 C 42.050000000000004 10.620000000000001 43.46 11.990000000000002 43.46 13.73 C 43.46 15.469999999999999 42.09 16.84 40.35 16.84 C 40.35 16.84 35.34 16.88 32.28 17.16 C 27.150000000000002 17.68 23.64 19.54 23.64 19.54 C 22.150000000000002 20.349999999999998 20.25 19.74 19.48 18.25 C 18.67 16.76 19.28 14.86 20.77 14.09 C 22.259999999999998 13.32 25.33 11.67 31.67 11.06 C 35.34 10.66 40.35 10.620000000000001 40.35 10.620000000000001 Z M 37.36 25.880000000000003 C 39.06 25.840000000000003 40.35 25.880000000000003 40.35 25.880000000000003 C 42.050000000000004 26.080000000000002 43.260000000000005 27.62 43.050000000000004 29.310000000000002 C 42.88374644848126 30.726609090871516 41.76660909087151 31.843746448481262 40.35 32.010000000000005 C 40.35 32.010000000000005 35.34 32.050000000000004 32.28 32.330000000000005 C 27.150000000000002 32.85000000000001 23.64 34.71000000000001 23.64 34.71000000000001 C 22.150000000000002 35.52000000000001 20.25 34.91000000000001 19.48 33.42000000000001 C 18.67 31.93000000000001 19.28 30.03000000000001 20.77 29.26000000000001 C 20.77 29.26000000000001 25.33 26.84000000000001 31.67 26.230000000000008 C 33.53 25.99000000000001 35.67 25.910000000000007 37.36 25.870000000000008 Z M 40.35 41.06 C 42.050000000000004 41.06 43.46 42.43 43.46 44.17 C 43.46 45.910000000000004 42.09 47.28 40.35 47.28 C 40.35 47.28 35.34 47.24 32.28 47.56 C 27.150000000000002 48.080000000000005 23.64 49.940000000000005 23.64 49.940000000000005 C 22.150000000000002 50.75000000000001 20.25 50.14000000000001 19.48 48.650000000000006 C 18.67 47.160000000000004 19.28 45.260000000000005 20.77 44.49000000000001 C 20.77 44.49000000000001 25.33 42.07000000000001 31.67 41.46000000000001 C 35.34 41.02000000000001 40.35 41.06000000000001 40.35 41.06000000000001 Z" stroke-linecap="round" />
+</g>
+<g transform="matrix(0.07 0 0 0.07 396.05 123.14)" style="" id="eb0df536-c517-4781-a7c0-3f84cd77c272"  >
+		<text xml:space="preserve" font-family="Lato" font-size="40" font-style="normal" font-weight="400" style="stroke: none; stroke-width: 1; stroke-dasharray: none; stroke-linecap: butt; stroke-dashoffset: 0; stroke-linejoin: miter; stroke-miterlimit: 4; fill: rgb(0,0,0); fill-rule: nonzero; opacity: 1; white-space: pre;" ><tspan x="-130" y="12.57" >Read The Docs</tspan></text>
+</g>
+<g transform="matrix(0.28 0 0 0.28 27.88 36)" id="7b9eddb9-1652-4040-9437-2ab90652d624"  >
+<path style="stroke: rgb(0,0,0); stroke-width: 0; stroke-dasharray: none; stroke-linecap: butt; stroke-dashoffset: 0; stroke-linejoin: miter; stroke-miterlimit: 4; fill: rgb(50,50,42); fill-rule: nonzero; opacity: 1;" vector-effect="non-scaling-stroke"  transform=" translate(-64, -64)" d="M 57.62 61.68 C 55.919999999999995 61.92 54.75 63.46 55 65.11 C 55.1668510745875 66.32380621250819 56.039448735907676 67.32218371690155 57.22 67.65 C 57.22 67.65 64.69 70.11 77.4 71.16000000000001 C 87.61000000000001 72.01 99.2 70.43 99.2 70.43 C 100.9 70.39 102.23 68.98 102.19 67.28 C 102.17037752125772 66.4652516996782 101.82707564255573 65.69186585376654 101.23597809465886 65.13079230830253 C 100.644880546762 64.56971876283853 99.85466451370849 64.26716220997277 99.03999999999999 64.29 C 98.83999999999999 64.29 98.63999999999999 64.33000000000001 98.42999999999999 64.37 C 98.42999999999999 64.37 87.08999999999999 65.78 77.88 65.02000000000001 C 65.72999999999999 64.05000000000001 59.11 61.83000000000001 59.11 61.83000000000001 C 58.63 61.670000000000016 58.1 61.59000000000001 57.62 61.670000000000016 Z M 57.62 46.46 C 55.919999999999995 46.7 54.75 48.24 55 49.89 C 55.1668510745875 51.10380621250818 56.039448735907676 52.10218371690154 57.22 52.43 C 57.22 52.43 64.69 54.89 77.4 55.94 C 87.61000000000001 56.79 99.2 55.21 99.2 55.21 C 100.9 55.17 102.23 53.76 102.19 52.06 C 102.17037752125772 51.245251699678214 101.82707564255573 50.47186585376654 101.23597809465886 49.91079230830253 C 100.644880546762 49.34971876283853 99.85466451370849 49.047162209972754 99.03999999999999 49.07 C 98.83999999999999 49.07 98.63999999999999 49.11 98.42999999999999 49.15 C 98.42999999999999 49.15 87.08999999999999 50.559999999999995 77.88 49.8 C 65.72999999999999 48.83 59.11 46.61 59.11 46.61 C 58.63 46.45 58.1 46.37 57.62 46.45 Z M 57.62 31.240000000000002 C 55.919999999999995 31.48 54.75 33.02 55 34.67 C 55.1668510745875 35.88380621250818 56.039448735907676 36.882183716901544 57.22 37.21 C 57.22 37.21 64.69 39.67 77.4 40.72 C 87.61000000000001 41.57 99.2 39.99 99.2 39.99 C 100.9 39.95 102.23 38.54 102.19 36.84 C 102.17037752125772 36.025251699678215 101.82707564255573 35.25186585376654 101.23597809465886 34.690792308302534 C 100.644880546762 34.12971876283853 99.85466451370849 33.827162209972755 99.03999999999999 33.85 C 98.83999999999999 33.85 98.63999999999999 33.89 98.42999999999999 33.93 C 98.42999999999999 33.93 87.08999999999999 35.339999999999996 77.88 34.58 C 65.72999999999999 33.61 59.11 31.389999999999997 59.11 31.389999999999997 C 58.63 31.229999999999997 58.1 31.189999999999998 57.62 31.229999999999997 Z M 57.62 16.060000000000002 C 55.919999999999995 16.3 54.75 17.840000000000003 55 19.490000000000002 C 55.1668510745875 20.703806212508187 56.039448735907676 21.702183716901544 57.22 22.03 C 57.22 22.03 64.69 24.490000000000002 77.4 25.54 C 87.61000000000001 26.39 99.2 24.81 99.2 24.81 C 100.9 24.77 102.23 23.36 102.19 21.66 C 102.17037752125772 20.84525169967821 101.82707564255573 20.07186585376654 101.23597809465886 19.510792308302534 C 100.644880546762 18.949718762838526 99.8546645137085 18.64716220997276 99.03999999999999 18.67 C 98.83999999999999 18.67 98.63999999999999 18.71 98.42999999999999 18.75 C 98.42999999999999 18.75 87.08999999999999 20.16 77.88 19.4 C 65.72999999999999 18.43 59.11 16.209999999999997 59.11 16.209999999999997 C 58.637850878541954 16.01924514007714 58.12188500879498 15.963839409097599 57.62 16.049999999999997 Z M 36.31 0 C 20.32 0.12 14.39 5.05 14.39 5.05 L 14.39 124.42 C 14.39 124.42 20.2 119.41 38.93 120.18 C 57.66 120.95000000000002 61.5 127.53 84.50999999999999 127.97000000000001 C 107.52 128.41000000000003 113.28999999999999 124.42000000000002 113.28999999999999 124.42000000000002 L 113.60999999999999 2.750000000000014 C 113.60999999999999 2.750000000000014 103.28 5.7 83.09 5.86 C 62.95 6.01 58.11 0.73 39.62 0.12 C 38.49 0.04 37.4 0 36.31 0 Z M 49.67 7.79 C 49.67 7.79 59.36 10.98 77.24000000000001 11.870000000000001 C 92.38000000000001 12.64 107.52000000000001 10.38 107.52000000000001 10.38 L 107.52000000000001 118.53 C 107.52000000000001 118.53 99.85000000000001 122.57000000000001 80.68 121.19 C 65.82000000000001 120.14 49.480000000000004 114.49 49.480000000000004 114.49 L 49.68000000000001 7.799999999999997 Z M 40.35 10.620000000000001 C 42.050000000000004 10.620000000000001 43.46 11.990000000000002 43.46 13.73 C 43.46 15.469999999999999 42.09 16.84 40.35 16.84 C 40.35 16.84 35.34 16.88 32.28 17.16 C 27.150000000000002 17.68 23.64 19.54 23.64 19.54 C 22.150000000000002 20.349999999999998 20.25 19.74 19.48 18.25 C 18.67 16.76 19.28 14.86 20.77 14.09 C 22.259999999999998 13.32 25.33 11.67 31.67 11.06 C 35.34 10.66 40.35 10.620000000000001 40.35 10.620000000000001 Z M 37.36 25.880000000000003 C 39.06 25.840000000000003 40.35 25.880000000000003 40.35 25.880000000000003 C 42.050000000000004 26.080000000000002 43.260000000000005 27.62 43.050000000000004 29.310000000000002 C 42.88374644848126 30.726609090871516 41.76660909087151 31.843746448481262 40.35 32.010000000000005 C 40.35 32.010000000000005 35.34 32.050000000000004 32.28 32.330000000000005 C 27.150000000000002 32.85000000000001 23.64 34.71000000000001 23.64 34.71000000000001 C 22.150000000000002 35.52000000000001 20.25 34.91000000000001 19.48 33.42000000000001 C 18.67 31.93000000000001 19.28 30.03000000000001 20.77 29.26000000000001 C 20.77 29.26000000000001 25.33 26.84000000000001 31.67 26.230000000000008 C 33.53 25.99000000000001 35.67 25.910000000000007 37.36 25.870000000000008 Z M 40.35 41.06 C 42.050000000000004 41.06 43.46 42.43 43.46 44.17 C 43.46 45.910000000000004 42.09 47.28 40.35 47.28 C 40.35 47.28 35.34 47.24 32.28 47.56 C 27.150000000000002 48.080000000000005 23.64 49.940000000000005 23.64 49.940000000000005 C 22.150000000000002 50.75000000000001 20.25 50.14000000000001 19.48 48.650000000000006 C 18.67 47.160000000000004 19.28 45.260000000000005 20.77 44.49000000000001 C 20.77 44.49000000000001 25.33 42.07000000000001 31.67 41.46000000000001 C 35.34 41.02000000000001 40.35 41.06000000000001 40.35 41.06000000000001 Z" stroke-linecap="round" />
+</g>
+<g transform="matrix(0.9 0 0 0.9 94 36)" style="" id="385bde16-f9fa-4222-bfea-1d5d5efcf730"  >
+		<text xml:space="preserve" font-family="Lato" font-size="15" font-style="normal" font-weight="100" style="stroke: none; stroke-width: 1; stroke-dasharray: none; stroke-linecap: butt; stroke-dashoffset: 0; stroke-linejoin: miter; stroke-miterlimit: 4; fill: rgb(0,0,0); fill-rule: nonzero; opacity: 1; white-space: pre;" ><tspan x="-48.68" y="4.71" >Read The Docs</tspan></text>
+</g>
+</svg>
\ No newline at end of file

next_docs/en/additional_notes/changelog.rst

-
-
-Changelog
-=========
-
--  2024/09/27 Version 0.8.1 released, Fixed some bugs, and providing a
-   `localized deployment version <projects/web_demo/README.md>`__ of the
-   `online
-   demo <https://opendatalab.com/OpenSourceTools/Extractor/PDF/>`__ and
-   the `front-end interface <projects/web/README.md>`__.
--  2024/09/09: Version 0.8.0 released, supporting fast deployment with
-   Dockerfile, and launching demos on Huggingface and Modelscope.
--  2024/08/30: Version 0.7.1 released, add paddle tablemaster table
-   recognition option
--  2024/08/09: Version 0.7.0b1 released, simplified installation
-   process, added table recognition functionality
--  2024/08/01: Version 0.6.2b1 released, optimized dependency conflict
-   issues and installation documentation
--  2024/07/05: Initial open-source release
-
-
-.. warning::
-
-   fix ``localized deployment version`` and ``front-end interface``
-
-

next_docs/en/additional_notes/faq.rst

@@ -74,3 +74,15 @@ CUDA version used by Paddle needs to be upgraded.
    pip install paddlepaddle-gpu==3.0.0b1 -i https://www.paddlepaddle.org.cn/packages/stable/cu123/
 
 Reference: https://github.com/opendatalab/MinerU/issues/558
+
+
+7. On some Linux servers, the program immediately reports an error ``Illegal instruction (core dumped)``
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+This might be because the server's CPU does not support the AVX/AVX2
+instruction set, or the CPU itself supports it but has been disabled by
+the system administrator. You can try contacting the system
+administrator to remove the restriction or change to a different server.
+
+References: https://github.com/opendatalab/MinerU/issues/591 ,
+https://github.com/opendatalab/MinerU/issues/736

next_docs/en/additional_notes/known_issues.rst

 Known Issues
 ============
 
--  Reading order is based on the model’s sorting of text distribution in
-   space, which may become disordered under extremely complex layouts.
+-  Reading order is determined by the model based on the spatial
+   distribution of readable content, and may be out of order in some
+   areas under extremely complex layouts.
 -  Vertical text is not supported.
--  Tables of contents and lists are recognized through rules; a few
-   uncommon list formats may not be identified.
--  Only one level of headings is supported; hierarchical heading levels
-   are currently not supported.
+-  Tables of contents and lists are recognized through rules, and some
+   uncommon list formats may not be recognized.
+-  Only one level of headings is supported; hierarchical headings are
+   not currently supported.
 -  Code blocks are not yet supported in the layout model.
--  Comic books, art books, elementary school textbooks, and exercise
-   books are not well-parsed yet
--  Enabling OCR may produce better results in PDFs with a high density
-   of formulas
--  If you are processing PDFs with a large number of formulas, it is
-   strongly recommended to enable the OCR function. When using PyMuPDF
-   to extract text, overlapping text lines can occur, leading to
-   inaccurate formula insertion positions.
+-  Comic books, art albums, primary school textbooks, and exercises
+   cannot be parsed well.
+-  Table recognition may result in row/column recognition errors in
+   complex tables.
+-  OCR recognition may produce inaccurate characters in PDFs of
+   lesser-known languages (e.g., diacritical marks in Latin script,
+   easily confused characters in Arabic script).
+-  Some formulas may not render correctly in Markdown.
\ No newline at end of file

next_docs/en/api.rst

@@ -7,4 +7,3 @@
    api/read_api
    api/schemas
    api/io
-   api/classes
\ No newline at end of file

next_docs/en/api/classes.rst

-Class Hierarchy
-===============
-
-.. inheritance-diagram:: magic_pdf.data.io.base magic_pdf.data.io.http magic_pdf.data.io.s3
-   :parts: 2
-
-
-.. inheritance-diagram:: magic_pdf.data.dataset
-   :parts: 2
-
-
-.. inheritance-diagram:: magic_pdf.data.data_reader_writer.base magic_pdf.data.data_reader_writer.filebase magic_pdf.data.data_reader_writer.multi_bucket_s3
-   :parts: 2
-

next_docs/en/api/utils.rst

-

next_docs/en/conf.py

@@ -95,7 +95,7 @@ language = 'en'
 html_theme = 'sphinx_book_theme'
 html_logo = '_static/image/logo.png'
 html_theme_options = {
-    'path_to_docs': 'docs/en',
+    'path_to_docs': 'next_docs/en',
     'repository_url': 'https://github.com/opendatalab/MinerU',
     'use_repository_button': True,
 }

next_docs/en/index.rst

@@ -46,20 +46,29 @@ the relevant PDF**.
 Key Features
 ------------
 
--  Removes elements such as headers, footers, footnotes, and page
-   numbers while maintaining semantic continuity
--  Outputs text in a human-readable order from multi-column documents
--  Retains the original structure of the document, including titles,
-   paragraphs, and lists
--  Extracts images, image captions, tables, and table captions
--  Automatically recognizes formulas in the document and converts them
-   to LaTeX
--  Automatically recognizes tables in the document and converts them to
-   LaTeX
--  Automatically detects and enables OCR for corrupted PDFs
--  Supports both CPU and GPU environments
--  Supports Windows, Linux, and Mac platforms
-
+-  Remove headers, footers, footnotes, page numbers, etc., to ensure
+   semantic coherence.
+-  Output text in human-readable order, suitable for single-column,
+   multi-column, and complex layouts.
+-  Preserve the structure of the original document, including headings,
+   paragraphs, lists, etc.
+-  Extract images, image descriptions, tables, table titles, and
+   footnotes.
+-  Automatically recognize and convert formulas in the document to LaTeX
+   format.
+-  Automatically recognize and convert tables in the document to LaTeX
+   or HTML format.
+-  Automatically detect scanned PDFs and garbled PDFs and enable OCR
+   functionality.
+-  OCR supports detection and recognition of 84 languages.
+-  Supports multiple output formats, such as multimodal and NLP
+   Markdown, JSON sorted by reading order, and rich intermediate
+   formats.
+-  Supports various visualization results, including layout
+   visualization and span visualization, for efficient confirmation of
+   output quality.
+-  Supports both CPU and GPU environments.
+-  Compatible with Windows, Linux, and Mac platforms.
 
 User Guide
 -------------
@@ -91,14 +100,6 @@ Additional Notes
 
    additional_notes/known_issues
    additional_notes/faq
-   additional_notes/changelog
    additional_notes/glossary
 
 
-Projects 
----------
-.. toctree::
-   :maxdepth: 1
-   :caption: Projects
-
-   projects
\ No newline at end of file

next_docs/en/projects.rst

-
-
-
-llama_index_rag 
-===============
-
-
-gradio_app
-============
-
-
-other projects
-===============
\ No newline at end of file

next_docs/en/user_guide/data/data_reader_writer.rst

@@ -87,6 +87,8 @@ Read Examples
 
 .. code:: python
 
+    from magic_pdf.data.data_reader_writer import *
+
     # file based related 
     file_based_reader1 = FileBasedDataReader('')
 
@@ -142,6 +144,8 @@ Write Examples
 
 .. code:: python
 
+    from magic_pdf.data.data_reader_writer import *
+
     # file based related 
     file_based_writer1 = FileBasedDataWriter('')
 
@@ -201,4 +205,4 @@ Write Examples
     s3_writer1.write('s3://test_bucket/efg', '123'.encode())
 
 
-Check :doc:`../../api/classes` for more intuitions or check :doc:`../../api/data_reader_writer` for more details
+Check :doc:`../../api/data_reader_writer` for more details

next_docs/en/user_guide/data/dataset.rst

@@ -36,5 +36,5 @@ Extract chars via third-party library, currently we use ``pymupdf``.
 
 
 
-Check :doc:`../../api/classes` for more intuitions or check :doc:`../../api/dataset` for more details
+Check :doc:`../../api/dataset` for more details
 

next_docs/en/user_guide/data/io.rst

@@ -21,5 +21,5 @@ if MinerU have not provide the suitable classes. It is easy to implement new cla
         def write(self, path: str, data: bytes) -> None:
             pass
 
-Check :doc:`../../api/classes` for more intuitions or check :doc:`../../api/io` for more details
+Check :doc:`../../api/io` for more details