{"id":2143,"date":"2026-08-21T10:40:06","date_gmt":"2026-08-21T02:40:06","guid":{"rendered":"https:\/\/www.puzi7.net\/?p=2143"},"modified":"2026-08-21T10:40:07","modified_gmt":"2026-08-21T02:40:07","slug":"77k-star-%e5%bc%80%e6%ba%90-pdf-%e8%a7%a3%e6%9e%90%e5%bc%95%e6%93%8e%ef%bc%9amineru-%e7%9a%84%e4%b8%89%e5%90%8e%e7%ab%af%e6%9e%b6%e6%9e%84%e4%b8%8e%e8%a1%a8%e6%a0%bc%e8%af%86%e5%88%ab%e6%b7%b1","status":"publish","type":"post","link":"https:\/\/www.puzi7.net\/?p=2143","title":{"rendered":"77K Star \u5f00\u6e90 PDF \u89e3\u6790\u5f15\u64ce\uff1aMinerU \u7684\u4e09\u540e\u7aef\u67b6\u6784\u4e0e\u8868\u683c\u8bc6\u522b\u6df1\u5ea6\u8bc4\u6d4b"},"content":{"rendered":"\n

PDF \u89e3\u6790\u662f RAG \u7ba1\u9053\u4e2d\u5de5\u7a0b\u91cf\u6700\u5bc6\u96c6\u7684\u73af\u8282\u3002PyMuPDF \u63d0\u53d6\u6587\u672c\u4f46\u4e22\u5931\u7ed3\u6784\uff0cunstructured \u505a\u7248\u9762\u5206\u6790\u4f46\u8868\u683c\u8bc6\u522b\u7c97\u7cd9\uff0c\u5546\u4e1a API \u8d28\u91cf\u597d\u4f46\u6210\u672c\u4e0d\u53ef\u63a7\u3002OpenDataLab \u7684 MinerU\uff08GitHub 77K Star\uff09\u8bd5\u56fe\u7528\u4e09\u540e\u7aef\u89e3\u8026\u67b6\u6784\u89e3\u51b3\u8fd9\u4e2a\u95ee\u9898\u3002<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

\u6280\u672f\u5b9a\u4f4d\uff1aLLM \u6570\u636e\u7ba1\u9053\u7684\u524d\u7f6e\u57fa\u7840\u8bbe\u65bd<\/h3>\n\n\n\n

MinerU \u4e0d\u662f\u53c8\u4e00\u4e2a PDF \u8f6c\u6587\u672c\u5de5\u5177\u3002\u5b83\u662f\u4e00\u4e2a\u9ad8\u7cbe\u5ea6\u6587\u6863\u89e3\u6790\u5f15\u64ce\uff0c\u5c06 PDF\u3001DOCX\u3001PPTX\u3001XLSX\u3001\u56fe\u7247\u8f6c\u6362\u4e3a\u7ed3\u6784\u5316\u7684 Markdown \u548c JSON\uff0c\u6838\u5fc3\u80fd\u529b\u662f\u7248\u9762\u7406\u89e3\u3001\u516c\u5f0f\u8f6c LaTeX\u3001\u8868\u683c\u8f6c HTML\u3001\u9605\u8bfb\u987a\u5e8f\u8fd8\u539f\u3002\u4e0e PyMuPDF \u7684\u672c\u8d28\u533a\u522b\u5728\u4e8e\uff1aMinerU \u505a\u7684\u662f\u6a21\u578b\u7ea7\u7684\u6587\u6863\u7ed3\u6784\u7406\u89e3\uff0c\u800c\u975e\u5b57\u7b26\u7ea7\u7684\u6587\u672c\u63d0\u53d6\u3002<\/p>\n\n\n\n

\u6838\u5fc3\u80fd\u529b\uff1a\u56db\u4e2a\u5de5\u7a0b\u5316\u8bbe\u8ba1<\/h3>\n\n\n\n

\u4e09\u540e\u7aef\u89e3\u8026\uff0c\u7cbe\u5ea6\u4e0e\u6210\u672c\u53ef\u8c03\u3002<\/strong> pipeline \u540e\u7aef\u7eaf CPU \u53ef\u8dd1\u3001\u65e0\u5e7b\u89c9\uff0c\u9002\u5408\u5feb\u901f\u5904\u7406\uff1bvlm-engine \u7528 MinerU2.5-Pro \u89c6\u89c9\u8bed\u8a00\u6a21\u578b\uff0c\u7cbe\u5ea6\u6700\u9ad8\uff1bhybrid-engine \u7ed3\u5408\u539f\u751f\u6587\u672c\u63d0\u53d6\u4e0e VLM \u8865\u5145\u3002OmniDocBench v1.6 \u57fa\u51c6\u4e0a\uff0cpipeline \u5f97\u5206 86.47\uff0chybrid\/vlm \u8fbe 95.30 \u4ee5\u4e0a\u30023.3 \u7248\u672c\u65b0\u589e effort \u53c2\u6570\uff0cmedium \u6a21\u5f0f\u76f8\u6bd4 high \u4ec5\u964d\u4f4e 0.13 \u5206\uff0c\u901f\u5ea6\u63d0\u5347 35%-220%\u3002<\/p>\n\n\n\n

\u8868\u683c\u8bc6\u522b\u662f\u6a21\u578b\u7ea7\u7ed3\u6784\u7406\u89e3\u3002<\/strong> \u4f7f\u7528 TableStructureRec \u8fdb\u884c\u8868\u683c\u7ed3\u6784\u8bc6\u522b\uff0c\u8f93\u51fa HTML \u683c\u5f0f\uff0c\u652f\u6301 rowspan\/colspan \u5408\u5e76\u5355\u5143\u683c\u30023.1.0 \u7248\u672c\u52a0\u5165\u8de8\u9875\u8868\u683c\u81ea\u52a8\u5408\u5e76\u548c\u8868\u683c\u5185\u56fe\u7247\/\u516c\u5f0f\u8bc6\u522b\uff0c\u81ea\u52a8\u63d0\u53d6\u8868\u683c\u6807\u9898\u548c\u811a\u6ce8\u3002\u8fd9\u4e0d\u662f OCR \u540e\u6b63\u5219\u5339\u914d\u7684\u65b9\u6848\u3002<\/p>\n\n\n\n

\u516c\u5f0f\u4e0e\u591a\u8bed\u8a00 OCR \u539f\u751f\u652f\u6301\u3002<\/strong> UniMERNet \u5c06\u6570\u5b66\u516c\u5f0f\u8f6c\u4e3a LaTeX\uff0cPP-OCRv6 \u652f\u6301 109 \u79cd\u8bed\u8a00\u8bc6\u522b\uff0c3.4 \u7248\u672c OCR \u51c6\u786e\u7387\u63d0\u5347\u7ea6 11%\u3002\u81ea\u52a8\u68c0\u6d4b\u626b\u63cf PDF \u548c\u4e71\u7801 PDF \u5e76\u542f\u7528 OCR\uff0c\u65e0\u9700\u624b\u52a8\u914d\u7f6e\u3002<\/p>\n\n\n\n

\u751f\u4ea7\u7ea7\u7f16\u6392\u67b6\u6784\u3002<\/strong> 3.0 \u7248\u672c\u5f15\u5165 mineru\uff08\u5ba2\u6237\u7aef\uff09+ mineru-api\uff08FastAPI \u670d\u52a1\uff09+ mineru-router\uff08\u591a GPU \u8d1f\u8f7d\u5747\u8861\uff09\u4e09\u5c42\u67b6\u6784\u3002\u7ebf\u7a0b\u5b89\u5168\u3001\u6ed1\u52a8\u7a97\u53e3\u5904\u7406\u957f\u6587\u6863\u3001\u6d41\u5f0f\u5199\u5165\u78c1\u76d8\u3002\u652f\u6301 Docker\u3001CLI\u3001REST API\u3001Gradio WebUI\u3001MCP Server\uff0c\u4ee5\u53ca Python\/Go\/TypeScript \u4e09\u8bed\u8a00 SDK\u3002<\/p>\n\n\n\n

\u98ce\u9669\u4e0e\u5c40\u9650<\/h3>\n\n\n\n

\ud83d\udfe1 \u8bb8\u53ef\u8bc1\u9700\u6cd5\u52a1\u5ba1\u67e5\u3002<\/strong> MinerU Open Source License \u57fa\u4e8e Apache 2.0 \u4f46\u9644\u52a0\u4e86\u6761\u4ef6\uff0cspdx_id \u4e3a NOASSERTION\u3002\u76f8\u6bd4 v3.1.0 \u4e4b\u524d\u7684 AGPLv3 \u5df2\u5927\u5e45\u6539\u5584\uff0c\u4f46\u5546\u4e1a\u90e8\u7f72\u524d\u4ecd\u9700\u5ba1\u67e5\u5177\u4f53\u6761\u6b3e\u3002\u9879\u76ee\u5df2\u5728 3.0.0 \u79fb\u9664\u4e86\u6240\u6709 AGPLv3 \u548c CC-BY-NC-SA \u4f9d\u8d56\u6a21\u578b\u3002<\/p>\n\n\n\n

\ud83d\udfe1 \u8d44\u6e90\u95e8\u69db\u4e0d\u4f4e\u3002<\/strong> pipeline \u540e\u7aef\u6700\u4f4e 4GB VRAM + 16GB RAM + 20GB \u78c1\u76d8\uff0cVLM \u540e\u7aef\u9700 8GB VRAM\u3002\u7eaf CPU \u53ef\u8fd0\u884c\u4f46\u901f\u5ea6\u6709\u9650\uff0c\u751f\u4ea7\u73af\u5883\u5efa\u8bae GPU \u52a0\u901f\u3002Windows \u56e0 ray \u4f9d\u8d56\u4e0d\u652f\u6301 Python 3.13\uff0cDocker \u4ec5\u652f\u6301 Linux\/WSL2\u3002<\/p>\n\n\n\n

\ud83d\udfe1 VLM \u540e\u7aef\u8f93\u51fa\u4e0e pipeline \u4e0d\u5411\u540e\u517c\u5bb9\u3002<\/strong> 2.5 \u7248\u672c VLM \u8f93\u51fa\u7ed3\u6784\u6709\u91cd\u5927\u53d8\u66f4\uff0c\u57fa\u4e8e\u7ed3\u6784\u5316\u8f93\u51fa\u505a\u4e8c\u6b21\u5f00\u53d1\u7684\u56e2\u961f\u9700\u8981\u9002\u914d\u3002content_list_v2.json \u662f 3.0 \u65b0\u589e\u7684\u7edf\u4e00\u683c\u5f0f\uff0c\u4f46\u4ecd\u662f\u5f00\u53d1\u7248\u672c\u3002<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

\u9009\u578b\u5efa\u8bae<\/h3>\n\n\n\n

\u63a8\u8350\u91c7\u7528\uff1a<\/strong> RAG \u77e5\u8bc6\u5e93\u6784\u5efa\u3001LLM \u6570\u636e\u9884\u5904\u7406\u3001\u5b66\u672f\u6587\u732e\/\u8d22\u62a5\/\u5408\u540c\u6279\u91cf\u89e3\u6790\u3001\u9700\u8981\u8868\u683c\u548c\u516c\u5f0f\u7ed3\u6784\u5316\u8f93\u51fa\u7684\u573a\u666f\u3001\u9700\u8981\u672c\u5730\u90e8\u7f72\u907f\u514d\u6570\u636e\u5916\u6d41\u7684\u4f01\u4e1a\u3002<\/p>\n\n\n\n

\u4e0d\u63a8\u8350\u91c7\u7528\uff1a<\/strong> \u6beb\u79d2\u7ea7\u5b9e\u65f6\u5728\u7ebf\u89e3\u6790\uff08VLM \u63a8\u7406\u6709\u5ef6\u8fdf\uff09\u3001\u7eaf CPU \u73af\u5883\u7684\u8d85\u5927\u6279\u91cf\u5904\u7406\uff08\u901f\u5ea6\u53d7\u9650\uff09\u3001\u5bf9\u8f93\u51fa\u683c\u5f0f\u7a33\u5b9a\u6027\u8981\u6c42\u6781\u9ad8\u7684\u4e8c\u6b21\u5f00\u53d1\uff08VLM \u8f93\u51fa\u4ecd\u5728\u8fed\u4ee3\uff09\u3002<\/p>\n\n\n\n

\u7ade\u54c1\u53d6\u820d\uff1a<\/strong> \u76f8\u6bd4 PyMuPDF\/pypdf\uff0cMinerU \u7684\u4f18\u52bf\u662f\u7ed3\u6784\u5316\u7406\u89e3\uff08\u8868\u683c\/\u516c\u5f0f\/\u7248\u9762\uff09\uff0c\u52a3\u52bf\u662f\u90e8\u7f72\u590d\u6742\u5ea6\u3002\u76f8\u6bd4 unstructured\uff0cMinerU \u7684\u8868\u683c\u8bc6\u522b\u7cbe\u5ea6\u548c\u4e2d\u6587\u652f\u6301\u663e\u8457\u66f4\u597d\u3002\u76f8\u6bd4\u5546\u4e1a API\uff08\u5982 Azure Document Intelligence\uff09\uff0cMinerU \u53ef\u672c\u5730\u90e8\u7f72\u3001\u65e0\u6570\u636e\u5916\u6d41\u3001\u65e0 API \u6210\u672c\uff0c\u4f46\u9700\u8981\u81ea\u884c\u8fd0\u7ef4\u3002<\/p>\n\n\n\n

\u843d\u5730\u5b9e\u8df5<\/h3>\n\n\n\n

\u5b89\u88c5\u63a8\u8350 pip install uv && uv pip install -U \"mineru[all]\"<\/code>\uff0c\u9996\u6b21\u4f7f\u7528\u9ed8\u8ba4\u4ece HuggingFace \u62c9\u53d6\u6a21\u578b\uff0c\u56fd\u5185\u73af\u5883\u8bbe export MINERU_MODEL_SOURCE=modelscope<\/code>\u3002\u5feb\u901f\u9a8c\u8bc1\u7528 mineru -p input.pdf -o output\/<\/code>\uff0cCPU \u73af\u5883\u52a0 -b pipeline<\/code>\u3002\u751f\u4ea7\u90e8\u7f72\u7528 mineru-api<\/code> \u542f\u52a8 FastAPI \u670d\u52a1\uff0c\u591a GPU \u573a\u666f\u7528 mineru-router --local-gpus auto<\/code>\u3002\u8868\u683c\u5bc6\u96c6\u578b\u6587\u6863\u5efa\u8bae\u7528 hybrid-engine + effort high\u3002\u957f\u6587\u6863\u65e0\u9700\u624b\u52a8\u62c6\u5206\uff0c\u6ed1\u52a8\u7a97\u53e3\u673a\u5236\u81ea\u52a8\u5904\u7406\u3002<\/p>\n\n\n\n

Github\u5730\u5740\uff1ahttps:\/\/github.com\/opendatalab\/MinerU<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"

PDF \u89e3\u6790\u662f RAG \u7ba1\u9053\u4e2d\u5de5\u7a0b\u91cf\u6700\u5bc6\u96c6\u7684\u73af\u8282\u3002PyMuPDF \u63d0\u53d6\u6587\u672c\u4f46\u4e22\u5931\u7ed3\u6784\uff0cun…<\/span><\/p>\n

Read More<\/a><\/p>\n","protected":false},"author":1,"featured_media":2144,"comment_status":"open","ping_status":"0","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[17,148,9],"tags":[165,40,168,149],"class_list":["post-2143","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai","category-skills","category-9","tag-agent","tag-ai","tag-mineru","tag-skills","zoom-theme-has-thumb"],"_links":{"self":[{"href":"https:\/\/www.puzi7.net\/index.php?rest_route=\/wp\/v2\/posts\/2143","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.puzi7.net\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.puzi7.net\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.puzi7.net\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.puzi7.net\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=2143"}],"version-history":[{"count":1,"href":"https:\/\/www.puzi7.net\/index.php?rest_route=\/wp\/v2\/posts\/2143\/revisions"}],"predecessor-version":[{"id":2147,"href":"https:\/\/www.puzi7.net\/index.php?rest_route=\/wp\/v2\/posts\/2143\/revisions\/2147"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.puzi7.net\/index.php?rest_route=\/wp\/v2\/media\/2144"}],"wp:attachment":[{"href":"https:\/\/www.puzi7.net\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=2143"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.puzi7.net\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=2143"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.puzi7.net\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=2143"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}