mineru 2.2.2__py3-none-any.whl → 2.5.0__py3-none-any.whl

mineru/cli/common.py

@@ -11,13 +11,14 @@ from loguru import logger
 from mineru.data.data_reader_writer import FileBasedDataWriter
 from mineru.utils.draw_bbox import draw_layout_bbox, draw_span_bbox, draw_line_sort_bbox
 from mineru.utils.enum_class import MakeMode
+from mineru.utils.guess_suffix_or_lang import guess_suffix_by_bytes
 from mineru.utils.pdf_image_tools import images_bytes_to_pdf_bytes
 from mineru.backend.vlm.vlm_middle_json_mkcontent import union_make as vlm_union_make
 from mineru.backend.vlm.vlm_analyze import doc_analyze as vlm_doc_analyze
 from mineru.backend.vlm.vlm_analyze import aio_doc_analyze as aio_vlm_doc_analyze
 
-pdf_suffixes = [".pdf"]
-image_suffixes = [".png", ".jpeg", ".jpg", ".webp", ".gif"]
+pdf_suffixes = ["pdf"]
+image_suffixes = ["png", "jpeg", "jp2", "webp", "gif", "bmp", "jpg"]
 
 
 def read_fn(path):
@@ -25,12 +26,13 @@ def read_fn(path):
         path = Path(path)
     with open(str(path), "rb") as input_file:
         file_bytes = input_file.read()
-        if path.suffix in image_suffixes:
+        file_suffix = guess_suffix_by_bytes(file_bytes)
+        if file_suffix in image_suffixes:
             return images_bytes_to_pdf_bytes(file_bytes)
-        elif path.suffix in pdf_suffixes:
+        elif file_suffix in pdf_suffixes:
             return file_bytes
         else:
-            raise Exception(f"Unknown file suffix: {path.suffix}")
+            raise Exception(f"Unknown file suffix: {file_suffix}")
 
 
 def prepare_env(output_dir, pdf_file_name, parse_method):
@@ -145,17 +147,10 @@ def _process_output(
         )
 
     if f_dump_model_output:
-        if is_pipeline:
-            md_writer.write_string(
-                f"{pdf_file_name}_model.json",
-                json.dumps(model_output, ensure_ascii=False, indent=4),
-            )
-        else:
-            output_text = ("\n" + "-" * 50 + "\n").join(model_output)
-            md_writer.write_string(
-                f"{pdf_file_name}_model_output.txt",
-                output_text,
-            )
+        md_writer.write_string(
+            f"{pdf_file_name}_model.json",
+            json.dumps(model_output, ensure_ascii=False, indent=4),
+        )
 
     logger.info(f"local output dir is {local_md_dir}")
 
@@ -333,6 +328,9 @@ def do_parse(
         if backend.startswith("vlm-"):
             backend = backend[4:]
 
+        if backend == "vllm-async-engine":
+            raise Exception("vlm-vllm-async-engine backend is not supported in sync mode, please use vlm-vllm-engine backend")
+
         os.environ['MINERU_VLM_FORMULA_ENABLE'] = str(formula_enable)
         os.environ['MINERU_VLM_TABLE_ENABLE'] = str(table_enable)
 
@@ -381,6 +379,9 @@ async def aio_do_parse(
         if backend.startswith("vlm-"):
             backend = backend[4:]
 
+        if backend == "vllm-engine":
+            raise Exception("vlm-vllm-engine backend is not supported in async mode, please use vlm-vllm-async-engine backend")
+
         os.environ['MINERU_VLM_FORMULA_ENABLE'] = str(formula_enable)
         os.environ['MINERU_VLM_TABLE_ENABLE'] = str(table_enable)
 

mineru/cli/fast_api.py

@@ -18,6 +18,7 @@ from base64 import b64encode
 
 from mineru.cli.common import aio_do_parse, read_fn, pdf_suffixes, image_suffixes
 from mineru.utils.cli_parser import arg_parse
+from mineru.utils.guess_suffix_or_lang import guess_suffix_by_path
 from mineru.version import __version__
 
 app = FastAPI()
@@ -95,13 +96,14 @@ async def parse_pdf(
             content = await file.read()
             file_path = Path(file.filename)
 
-            # 如果是图像文件或PDF，使用read_fn处理
-            if file_path.suffix.lower() in pdf_suffixes + image_suffixes:
-                # 创建临时文件以便使用read_fn
-                temp_path = Path(unique_dir) / file_path.name
-                with open(temp_path, "wb") as f:
-                    f.write(content)
+            # 创建临时文件
+            temp_path = Path(unique_dir) / file_path.name
+            with open(temp_path, "wb") as f:
+                f.write(content)
 
+            # 如果是图像文件或PDF，使用read_fn处理
+            file_suffix = guess_suffix_by_path(temp_path)
+            if file_suffix in pdf_suffixes + image_suffixes:
                 try:
                     pdf_bytes = read_fn(temp_path)
                     pdf_bytes_list.append(pdf_bytes)
@@ -115,7 +117,7 @@ async def parse_pdf(
             else:
                 return JSONResponse(
                     status_code=400,
-                    content={"error": f"Unsupported file type: {file_path.suffix}"}
+                    content={"error": f"Unsupported file type: {file_suffix}"}
                 )
 
 

mineru/cli/gradio_app.py

@@ -182,9 +182,9 @@ def to_pdf(file_path):
 
 # 更新界面函数
 def update_interface(backend_choice):
-    if backend_choice in ["vlm-transformers", "vlm-sglang-engine"]:
+    if backend_choice in ["vlm-transformers", "vlm-vllm-async-engine"]:
         return gr.update(visible=False), gr.update(visible=False)
-    elif backend_choice in ["vlm-sglang-client"]:
+    elif backend_choice in ["vlm-http-client"]:
         return gr.update(visible=True), gr.update(visible=False)
     elif backend_choice in ["pipeline"]:
         return gr.update(visible=False), gr.update(visible=True)
@@ -203,10 +203,10 @@ def update_interface(backend_choice):
     default=True,
 )
 @click.option(
-    '--enable-sglang-engine',
-    'sglang_engine_enable',
+    '--enable-vllm-engine',
+    'vllm_engine_enable',
     type=bool,
-    help="Enable SgLang engine backend for faster processing.",
+    help="Enable vLLM engine backend for faster processing.",
     default=False,
 )
 @click.option(
@@ -246,7 +246,7 @@ def update_interface(backend_choice):
     default='all',
 )
 def main(ctx,
-        example_enable, sglang_engine_enable, api_enable, max_convert_pages,
+        example_enable, vllm_engine_enable, api_enable, max_convert_pages,
         server_name, server_port, latex_delimiters_type, **kwargs
 ):
 
@@ -261,22 +261,21 @@ def main(ctx,
     else:
         raise ValueError(f"Invalid latex delimiters type: {latex_delimiters_type}.")
 
-    if sglang_engine_enable:
+    if vllm_engine_enable:
         try:
-            print("Start init SgLang engine...")
+            print("Start init vLLM engine...")
             from mineru.backend.vlm.vlm_analyze import ModelSingleton
             model_singleton = ModelSingleton()
             predictor = model_singleton.get_model(
-                "sglang-engine",
+                "vllm-async-engine",
                 None,
                 None,
                 **kwargs
             )
-            print("SgLang engine init successfully.")
+            print("vLLM engine init successfully.")
         except Exception as e:
             logger.exception(e)
-
-    suffixes = pdf_suffixes + image_suffixes
+    suffixes = [f".{suffix}" for suffix in pdf_suffixes + image_suffixes]
     with gr.Blocks() as demo:
         gr.HTML(header)
         with gr.Row():
@@ -286,11 +285,11 @@ def main(ctx,
                 with gr.Row():
                     max_pages = gr.Slider(1, max_convert_pages, int(max_convert_pages/2), step=1, label='Max convert pages')
                 with gr.Row():
-                    if sglang_engine_enable:
-                        drop_list = ["pipeline", "vlm-sglang-engine"]
-                        preferred_option = "vlm-sglang-engine"
+                    if vllm_engine_enable:
+                        drop_list = ["pipeline", "vlm-vllm-async-engine"]
+                        preferred_option = "vlm-vllm-async-engine"
                     else:
-                        drop_list = ["pipeline", "vlm-transformers", "vlm-sglang-client"]
+                        drop_list = ["pipeline", "vlm-transformers", "vlm-http-client"]
                         preferred_option = "pipeline"
                     backend = gr.Dropdown(drop_list, label="Backend", value=preferred_option)
                 with gr.Row(visible=False) as client_options:

mineru/cli/vlm_vllm_server.py

@@ -0,0 +1,4 @@
+from mineru.model.vlm_vllm_model.server import main
+
+if __name__ == "__main__":
+    main()

mineru/model/table/rec/unet_table/main.py

@@ -12,6 +12,7 @@ from PIL import Image
 from loguru import logger
 from bs4 import BeautifulSoup
 
+from mineru.utils.span_pre_proc import calculate_contrast
 from .table_structure_unet import TSRUnet
 
 from mineru.utils.enum_class import ModelPath
@@ -191,6 +192,13 @@ class WiredTableRecognition:
                 # logger.warning(f"Box {i} has invalid aspect ratio: {x1, y1, x2, y2}")
                 continue
             img_crop = bgr_img[int(y1):int(y2), int(x1):int(x2)]
+
+            # 计算span的对比度，低于0.20的span不进行ocr
+            if calculate_contrast(img_crop, img_mode='bgr') <= 0.17:
+                cell_box_map[i] = [[box, "", 0.1]]
+                # logger.debug(f"Box {i} skipped due to low contrast.")
+                continue
+
             img_crop_list.append(img_crop)
             img_crop_info_list.append([i, box])
 

mineru/model/vlm_vllm_model/server.py

@@ -0,0 +1,51 @@
+import sys
+
+from mineru.utils.models_download_utils import auto_download_and_get_model_root_path
+from vllm.entrypoints.cli.main import main as vllm_main
+
+
+def main():
+    args = sys.argv[1:]
+
+    has_port_arg = False
+    has_gpu_memory_utilization_arg = False
+    model_path = None
+    model_arg_indices = []
+
+    # 检查现有参数
+    for i, arg in enumerate(args):
+        if arg == "--port" or arg.startswith("--port="):
+            has_port_arg = True
+        if arg == "--gpu-memory-utilization" or arg.startswith("--gpu-memory-utilization="):
+            has_gpu_memory_utilization_arg = True
+        if arg == "--model":
+            if i + 1 < len(args):
+                model_path = args[i + 1]
+                model_arg_indices.extend([i, i + 1])
+        elif arg.startswith("--model="):
+            model_path = arg.split("=", 1)[1]
+            model_arg_indices.append(i)
+
+    # 从参数列表中移除 --model 参数
+    if model_arg_indices:
+        for index in sorted(model_arg_indices, reverse=True):
+            args.pop(index)
+
+    # 添加默认参数
+    if not has_port_arg:
+        args.extend(["--port", "30000"])
+    if not has_gpu_memory_utilization_arg:
+        args.extend(["--gpu-memory-utilization", "0.5"])
+    if not model_path:
+        model_path = auto_download_and_get_model_root_path("/", "vlm")
+
+    # 重构参数，将模型路径作为位置参数
+    sys.argv = [sys.argv[0]] + ["serve", model_path] + args
+
+    # 启动vllm服务器
+    print(f"start vllm server: {sys.argv}")
+    vllm_main()
+
+
+if __name__ == "__main__":
+    main()

mineru/resources/header.html

@@ -54,7 +54,7 @@
             font-family: 'Trebuchet MS', 'Lucida Sans Unicode',
               'Lucida Grande', 'Lucida Sans', Arial, sans-serif;
           ">
-          MinerU 2: PDF Extraction Demo
+          MinerU 2.5: PDF Extraction Demo
         </h1>
       </div>
     </div>
@@ -88,7 +88,15 @@
             <span style="color: white">Code</span>
           </a>
         </span>
-
+        <!-- Code Link. -->
+        <span class="link-block">
+          <a href="https://huggingface.co/opendatalab/MinerU2.5-2509-1.2B" class="external-link button is-normal is-rounded is-dark" style="text-decoration: none; cursor: pointer">
+            <span class="icon" style="margin-right: 4px">
+              <i class="fas fa-archive" style="color: white; margin-right: 4px"></i>
+            </span>
+            <span style="color: white">Model</span>
+          </a>
+        </span>
         <!-- arXiv Link. -->
         <span class="link-block">
           <a href="https://arxiv.org/abs/2409.18839" class="external-link button is-normal is-rounded is-dark" style="text-decoration: none; cursor: pointer">

mineru/utils/draw_bbox.py

@@ -119,22 +119,26 @@ def draw_bbox_with_number(i, bbox_list, page, c, rgb_config, fill_config, draw_b
 
 def draw_layout_bbox(pdf_info, pdf_bytes, out_path, filename):
     dropped_bbox_list = []
-    tables_list, tables_body_list = [], []
-    tables_caption_list, tables_footnote_list = [], []
-    imgs_list, imgs_body_list, imgs_caption_list, imgs_footnote_list = [], [], [], []
+    tables_body_list, tables_caption_list, tables_footnote_list = [], [], []
+    imgs_body_list, imgs_caption_list, imgs_footnote_list = [], [], []
+    codes_body_list, codes_caption_list = [], []
     titles_list = []
     texts_list = []
     interequations_list = []
     lists_list = []
+    list_items_list = []
     indexs_list = []
+
     for page in pdf_info:
         page_dropped_list = []
-        tables, tables_body, tables_caption, tables_footnote = [], [], [], []
-        imgs, imgs_body, imgs_caption, imgs_footnote = [], [], [], []
+        tables_body, tables_caption, tables_footnote = [], [], []
+        imgs_body, imgs_caption, imgs_footnote = [], [], []
+        codes_body, codes_caption = [], []
         titles = []
         texts = []
         interequations = []
         lists = []
+        list_items = []
         indices = []
 
         for dropped_bbox in page['discarded_blocks']:
@@ -143,7 +147,6 @@ def draw_layout_bbox(pdf_info, pdf_bytes, out_path, filename):
         for block in page["para_blocks"]:
             bbox = block["bbox"]
             if block["type"] == BlockType.TABLE:
-                tables.append(bbox)
                 for nested_block in block["blocks"]:
                     bbox = nested_block["bbox"]
                     if nested_block["type"] == BlockType.TABLE_BODY:
@@ -155,7 +158,6 @@ def draw_layout_bbox(pdf_info, pdf_bytes, out_path, filename):
                             continue
                         tables_footnote.append(bbox)
             elif block["type"] == BlockType.IMAGE:
-                imgs.append(bbox)
                 for nested_block in block["blocks"]:
                     bbox = nested_block["bbox"]
                     if nested_block["type"] == BlockType.IMAGE_BODY:
@@ -164,22 +166,31 @@ def draw_layout_bbox(pdf_info, pdf_bytes, out_path, filename):
                         imgs_caption.append(bbox)
                     elif nested_block["type"] == BlockType.IMAGE_FOOTNOTE:
                         imgs_footnote.append(bbox)
+            elif block["type"] == BlockType.CODE:
+                for nested_block in block["blocks"]:
+                    if nested_block["type"] == BlockType.CODE_BODY:
+                        bbox = nested_block["bbox"]
+                        codes_body.append(bbox)
+                    elif nested_block["type"] == BlockType.CODE_CAPTION:
+                        bbox = nested_block["bbox"]
+                        codes_caption.append(bbox)
             elif block["type"] == BlockType.TITLE:
                 titles.append(bbox)
-            elif block["type"] == BlockType.TEXT:
+            elif block["type"] in [BlockType.TEXT, BlockType.REF_TEXT]:
                 texts.append(bbox)
             elif block["type"] == BlockType.INTERLINE_EQUATION:
                 interequations.append(bbox)
             elif block["type"] == BlockType.LIST:
                 lists.append(bbox)
+                if "blocks" in block:
+                    for sub_block in block["blocks"]:
+                        list_items.append(sub_block["bbox"])
             elif block["type"] == BlockType.INDEX:
                 indices.append(bbox)
 
-        tables_list.append(tables)
         tables_body_list.append(tables_body)
         tables_caption_list.append(tables_caption)
         tables_footnote_list.append(tables_footnote)
-        imgs_list.append(imgs)
         imgs_body_list.append(imgs_body)
         imgs_caption_list.append(imgs_caption)
         imgs_footnote_list.append(imgs_footnote)
@@ -187,7 +198,10 @@ def draw_layout_bbox(pdf_info, pdf_bytes, out_path, filename):
         texts_list.append(texts)
         interequations_list.append(interequations)
         lists_list.append(lists)
+        list_items_list.append(list_items)
         indexs_list.append(indices)
+        codes_body_list.append(codes_body)
+        codes_caption_list.append(codes_caption)
 
     layout_bbox_list = []
 
@@ -197,6 +211,7 @@ def draw_layout_bbox(pdf_info, pdf_bytes, out_path, filename):
         for block in page["para_blocks"]:
             if block["type"] in [
                 BlockType.TEXT,
+                BlockType.REF_TEXT,
                 BlockType.TITLE,
                 BlockType.INTERLINE_EQUATION,
                 BlockType.LIST,
@@ -215,6 +230,10 @@ def draw_layout_bbox(pdf_info, pdf_bytes, out_path, filename):
                         continue
                     bbox = sub_block["bbox"]
                     page_block_list.append(bbox)
+            elif block["type"] in [BlockType.CODE]:
+                for sub_block in block["blocks"]:
+                    bbox = sub_block["bbox"]
+                    page_block_list.append(bbox)
 
         layout_bbox_list.append(page_block_list)
 
@@ -231,6 +250,8 @@ def draw_layout_bbox(pdf_info, pdf_bytes, out_path, filename):
         # 使用原始PDF的尺寸创建canvas
         c = canvas.Canvas(packet, pagesize=custom_page_size)
 
+        c = draw_bbox_without_number(i, codes_body_list, page, c, [102, 0, 204], True)
+        c = draw_bbox_without_number(i, codes_caption_list, page, c, [204, 153, 255], True)
         c = draw_bbox_without_number(i, dropped_bbox_list, page, c, [158, 158, 158], True)
         c = draw_bbox_without_number(i, tables_body_list, page, c, [204, 204, 0], True)
         c = draw_bbox_without_number(i, tables_caption_list, page, c, [255, 255, 102], True)
@@ -242,6 +263,7 @@ def draw_layout_bbox(pdf_info, pdf_bytes, out_path, filename):
         c = draw_bbox_without_number(i, texts_list, page, c, [153, 0, 76], True)
         c = draw_bbox_without_number(i, interequations_list, page, c, [0, 255, 0], True)
         c = draw_bbox_without_number(i, lists_list, page, c, [40, 169, 92], True)
+        c = draw_bbox_without_number(i, list_items_list, page, c, [40, 169, 92], False)
         c = draw_bbox_without_number(i, indexs_list, page, c, [40, 169, 92], True)
         c = draw_bbox_with_number(i, layout_bbox_list, page, c, [255, 0, 0], False, draw_bbox=False)
 

mineru/utils/enum_class.py

@@ -14,6 +14,19 @@ class BlockType:
     INDEX = 'index'
     DISCARDED = 'discarded'
 
+    # Added in vlm 2.5
+    CODE = "code"
+    CODE_BODY = "code_body"
+    CODE_CAPTION = "code_caption"
+    ALGORITHM = "algorithm"
+    REF_TEXT = "ref_text"
+    PHONETIC = "phonetic"
+    HEADER = "header"
+    FOOTER = "footer"
+    PAGE_NUMBER = "page_number"
+    ASIDE_TEXT = "aside_text"
+    PAGE_FOOTNOTE = "page_footnote"
+
 
 class ContentType:
     IMAGE = 'image'
@@ -22,6 +35,7 @@ class ContentType:
     INTERLINE_EQUATION = 'interline_equation'
     INLINE_EQUATION = 'inline_equation'
     EQUATION = 'equation'
+    CODE = 'code'
 
 
 class CategoryId:
@@ -49,8 +63,8 @@ class MakeMode:
 
 
 class ModelPath:
-    vlm_root_hf = "opendatalab/MinerU2.0-2505-0.9B"
-    vlm_root_modelscope = "OpenDataLab/MinerU2.0-2505-0.9B"
+    vlm_root_hf = "opendatalab/MinerU2.5-2509-1.2B"
+    vlm_root_modelscope = "OpenDataLab/MinerU2.5-2509-1.2B"
     pipeline_root_modelscope = "OpenDataLab/PDF-Extract-Kit-1.0"
     pipeline_root_hf = "opendatalab/PDF-Extract-Kit-1.0"
     doclayout_yolo = "models/Layout/YOLO/doclayout_yolo_docstructbench_imgsz1280_2501.pt"

mineru/utils/guess_suffix_or_lang.py

@@ -0,0 +1,20 @@
+from magika import Magika
+
+
+DEFAULT_LANG = "txt"
+magika = Magika()
+
+def guess_language_by_text(code):
+    codebytes = code.encode(encoding="utf-8")
+    lang = magika.identify_bytes(codebytes).prediction.output.label
+    return lang if lang != "unknown" else DEFAULT_LANG
+
+
+def guess_suffix_by_bytes(file_bytes) -> str:
+    suffix = magika.identify_bytes(file_bytes).prediction.output.label
+    return suffix
+
+
+def guess_suffix_by_path(file_path) -> str:
+    suffix = magika.identify_path(file_path).prediction.output.label
+    return suffix

mineru/utils/span_block_fix.py

@@ -23,9 +23,11 @@ def fill_spans_in_blocks(blocks, spans, radio):
             block_dict['group_id'] = block[-1]
         block_spans = []
         for span in spans:
+            temp_radio = radio
             span_bbox = span['bbox']
-            if calculate_overlap_area_in_bbox1_area_ratio(span_bbox, block_bbox) > radio and span_block_type_compatible(
-                    span['type'], block_type):
+            if span['type'] in [ContentType.IMAGE, ContentType.TABLE]:
+                temp_radio = 0.9
+            if calculate_overlap_area_in_bbox1_area_ratio(span_bbox, block_bbox) > temp_radio and span_block_type_compatible(span['type'], block_type):
                 block_spans.append(span)
 
         block_dict['spans'] = block_spans

mineru/version.py

@@ -1 +1 @@
-__version__ = "2.2.2"
+__version__ = "2.5.0"