{"id":96,"date":"2026-07-02T19:03:42","date_gmt":"2026-07-02T11:03:42","guid":{"rendered":"https:\/\/qiwanyang.xyz\/blog\/?p=96"},"modified":"2026-07-02T19:21:11","modified_gmt":"2026-07-02T11:21:11","slug":"%e8%ae%a1%e7%ae%97%e6%9c%ba%e8%a7%86%e8%a7%89%e5%ae%9e%e8%ae%ad%ef%bc%9a%e5%9f%ba%e4%ba%8eyolov8n-pose%e7%9a%84%e8%84%8a%e6%9f%b1x%e5%85%89%e7%89%87%e8%af%86%e5%88%ab%e5%88%86%e6%9e%90%e7%b3%bb","status":"publish","type":"post","link":"https:\/\/qiwanyang.xyz\/blog\/index.php\/2026\/07\/02\/%e8%ae%a1%e7%ae%97%e6%9c%ba%e8%a7%86%e8%a7%89%e5%ae%9e%e8%ae%ad%ef%bc%9a%e5%9f%ba%e4%ba%8eyolov8n-pose%e7%9a%84%e8%84%8a%e6%9f%b1x%e5%85%89%e7%89%87%e8%af%86%e5%88%ab%e5%88%86%e6%9e%90%e7%b3%bb\/","title":{"rendered":"\u8ba1\u7b97\u673a\u89c6\u89c9\u5b9e\u8bad\uff1a\u57fa\u4e8eYOLOv8n-pose\u7684\u810a\u67f1X\u5149\u7247\u8bc6\u522b\u5206\u6790\u7cfb\u7edf"},"content":{"rendered":"\r\n

\u8ba1\u7b97\u673a\u89c6\u89c9\u5b9e\u8bad\uff1a\u57fa\u4e8eYOLOv8n-pose\u7684\u810a\u67f1X\u5149\u7247\u8bc6\u522b\u5206\u6790\u7cfb\u7edf<\/h1>\r\n\r\n\r\n\r\n
\u672c\u6b21\u5b9e\u8bad\u5b8c\u6574\u8bb0\u5f55\uff1a\u4ece\u5168\u73ed55\u4eba\u624b\u5de5\u6807\u6ce81300\u5f20\u6570\u636e\uff0c\u5230983\u5f20\u6709\u6548\u6570\u636e\u7684\u6e05\u6d17\uff0c\u518d\u5230YOLOv8n-pose\u6a21\u578b\u8bad\u7ec3\u3001\u4e34\u5e8a\u53c2\u6570\u81ea\u52a8\u8ba1\u7b97\u3001Gradio\u53ef\u89c6\u5316\u754c\u9762\u642d\u5efa\uff0c\u4ee5\u53ca\u5927\u6a21\u578bAPI\u751f\u6210\u533b\u5b66\u62a5\u544a\u7684\u5168\u6d41\u7a0b\u3002 <\/blockquote>\r\n\r\n\r\n
\r\n\r\n\r\n

\u4e00\u3001\u9879\u76ee\u80cc\u666f\u4e0e\u610f\u4e49<\/h2>\r\n\r\n\r\n\r\n

\u810a\u67f1\u4fa7\u5f2f\uff08Scoliosis\uff09\u662f\u4e00\u79cd\u5e38\u89c1\u7684\u810a\u67f1\u4e09\u7ef4\u7578\u5f62\u75be\u75c5\uff0c\u597d\u53d1\u4e8e\u9752\u5c11\u5e74\u7fa4\u4f53\u3002\u636e\u7edf\u8ba1\uff0c\u6211\u56fd\u4e2d\u5c0f\u5b66\u751f\u810a\u67f1\u4fa7\u5f2f\u53d1\u75c5\u7387\u7ea6\u4e3a1%~3%\uff0c\u5df2\u6210\u4e3a\u7ee7\u8fd1\u89c6\u3001\u80a5\u80d6\u4e4b\u540e\u7684\u7b2c\u4e09\u5927\u9752\u5c11\u5e74\u5065\u5eb7\u95ee\u9898\u3002\u4f20\u7edf\u7684\u810a\u67f1\u4fa7\u5f2f\u8bca\u65ad\u4f9d\u8d56\u533b\u751f\u624b\u5de5\u6d4b\u91cfX\u5149\u7247\u4e0a\u7684Cobb\u89d2\u7b49\u53c2\u6570\uff0c\u4e0d\u4ec5\u8017\u65f6\u8017\u529b\uff0c\u4e14\u4e0d\u540c\u533b\u751f\u4e4b\u95f4\u7684\u6d4b\u91cf\u7ed3\u679c\u5b58\u5728\u4e00\u5b9a\u4e3b\u89c2\u5dee\u5f02\u3002<\/p>\r\n\r\n\r\n\r\n

\u672c\u6b21\u5b9e\u8bad\u65e8\u5728\u63a2\u7d22\u57fa\u4e8e\u6df1\u5ea6\u5b66\u4e60\u7684\u810a\u67f1X\u5149\u7247\u81ea\u52a8\u8bc6\u522b\u4e0e\u5206\u6790\u7cfb\u7edf<\/strong>\uff0c\u901a\u8fc7\u8ba1\u7b97\u673a\u89c6\u89c9\u6280\u672f\u5b9e\u73b0\uff1a<\/p>\r\n\r\n\r\n\r\n

    \r\n
  • \u810a\u67f1\u5173\u952e\u70b9\u7684\u81ea\u52a8\u68c0\u6d4b\u4e0e\u5b9a\u4f4d\uff0818\u4e2a\u690e\u4f53\u4e2d\u5fc3\u70b9\uff09<\/li>\r\n
  • \u4e34\u5e8a\u53c2\u6570\uff08Cobb\u89d2\u3001C7PL-CSVL\u3001T1\u503e\u659c\u89d2\u7b49\uff09\u7684\u81ea\u52a8\u8ba1\u7b97<\/li>\r\n
  • \u7ed3\u6784\u5316\u533b\u5b66\u62a5\u544a\u7684\u81ea\u52a8\u751f\u6210\uff08\u63a5\u5165\u5927\u8bed\u8a00\u6a21\u578b\uff09<\/li>\r\n
  • Gradio\u53ef\u4ea4\u4e92Web\u754c\u9762\uff0c\u652f\u6301\u62d6\u62fd\u4e0a\u4f20\u3001\u5b9e\u65f6\u5206\u6790<\/li>\r\n<\/ul>\r\n\r\n\r\n\r\n

    \u8fd9\u4e00\u65b9\u5411\u5177\u6709\u91cd\u8981\u7684\u4e34\u5e8a\u5e94\u7528\u4ef7\u503c\uff0c\u80fd\u591f\u8f85\u52a9\u533b\u751f\u63d0\u9ad8\u8bca\u65ad\u6548\u7387\uff0c\u964d\u4f4e\u4e3b\u89c2\u8bef\u5dee\uff0c\u540c\u65f6\u4e3a\u57fa\u5c42\u533b\u7597\u673a\u6784\u63d0\u4f9b\u4f4e\u6210\u672c\u3001\u9ad8\u8d28\u91cf\u7684\u810a\u67f1\u7b5b\u67e5\u5de5\u5177\u3002<\/p>\r\n\r\n\r\n

    \r\n\r\n\r\n

    \u4e8c\u3001\u6570\u636e\u96c6\u6784\u5efa\uff1a\u5168\u73ed\u534f\u4f5c\u7684\u6807\u6ce8\u5de5\u7a0b<\/h2>\r\n\r\n\r\n\r\n

    2.1 \u6570\u636e\u6765\u6e90\u4e0e\u6807\u6ce8\u4efb\u52a1\u5206\u914d<\/h3>\r\n\r\n\r\n\r\n

    \u672c\u6b21\u5b9e\u8bad\u7684\u6570\u636e\u96c6\u6765\u6e90\u4e8e\u5168\u73ed55\u4f4d\u540c\u5b66\u7684\u534f\u4f5c\u6807\u6ce8<\/strong>\u3002\u6bcf\u4f4d\u540c\u5b66\u4f7f\u7528 PyCharm \u73af\u5883\u4e0b\u7684 LabelMe<\/strong> \u5de5\u5177\uff0c\u5bf9\u810a\u67f1X\u5149\u6b63\u4f4d\u7247\u8fdb\u884c\u5173\u952e\u70b9\u6807\u6ce8\uff0c\u6bcf\u4eba\u8d1f\u8d23\u7ea631\u5f20\u56fe\u7247\uff0c\u603b\u8ba1\u91c7\u96c6\u4e861300\u591a\u5f20<\/strong>\u539f\u59cb\u6807\u6ce8\u6570\u636e\u3002<\/p>\r\n\r\n\r\n\r\n

    \u6807\u6ce8\u89c4\u8303\uff1a<\/strong><\/p>\r\n\r\n\r\n\r\n

      \r\n
    • \u6bcf\u4e2a\u690e\u4f53\u6807\u6ce8 4\u4e2a\u89d2\u70b9<\/strong>\uff08\u5de6\u4e0a\u3001\u53f3\u4e0a\u3001\u5de6\u4e0b\u3001\u53f3\u4e0b\uff09\uff0c\u517118\u4e2a\u690e\u4f53<\/li>\r\n
    • \u690e\u4f53\u987a\u5e8f\uff1aC7 \u2192 T1 \u2192 T2 \u2192 … \u2192 T12 \u2192 L1 \u2192 L2 \u2192 … \u2192 L5<\/li>\r\n
    • \u6807\u7b7e\u6587\u4ef6\u4fdd\u5b58\u4e3a .mat<\/code> \u683c\u5f0f\uff0c\u5176\u4e2d points<\/code> \u5b57\u6bb5\u5b58\u50a8\u89d2\u70b9\u5750\u6807 (x, y)<\/code><\/li>\r\n<\/ul>\r\n

      \"\"<\/p>\r\n\r\n\r\n\r\n\r\n\r\n

      2.2 \u6570\u636e\u6e05\u6d17\uff1a\u4ece1300\u5f20\u5230983\u5f20<\/h3>\r\n\r\n\r\n\r\n

      \u539f\u59cb\u6570\u636e\u5b58\u5728\u8f83\u591a\u8d28\u91cf\u95ee\u9898\uff0c\u7ecf\u8fc7\u7cfb\u7edf\u6e05\u6d17\u540e\u5f97\u5230\u6709\u6548\u6570\u636e 983\u5f20<\/strong>\uff1a<\/p>\r\n\r\n\r\n\r\n

      \r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
      \u95ee\u9898\u7c7b\u578b<\/th>\r\n\u5177\u4f53\u60c5\u51b5<\/th>\r\n\u5904\u7406\u65b9\u5f0f<\/th>\r\n<\/tr>\r\n<\/thead>\r\n
      \u975e18\u8282\u810a\u67f1<\/td>\r\n\u7ea6320\u591a\u5f20\u56fe\u7247\u6807\u6ce8\u7684\u690e\u4f53\u6570\u91cf\u4e0d\u4e3a18<\/td>\r\n\u5254\u9664<\/td>\r\n<\/tr>\r\n
      \u56fe\u540d\u4e0d\u5339\u914d<\/td>\r\n\u90e8\u5206\u56fe\u7247\u6587\u4ef6\u540d\u4e0e\u6807\u7b7e\u6587\u4ef6\u540d\u4e0d\u4e00\u81f4<\/td>\r\n\u4eba\u5de5\u6838\u5bf9\u540e\u4fee\u590d\u6216\u5254\u9664<\/td>\r\n<\/tr>\r\n
      \u6807\u6ce8\u987a\u5e8f\u9519\u8bef<\/td>\r\n\u5c11\u6570\u540c\u5b66\u89d2\u70b9\u987a\u5e8f\u975e”\u5de6\u4e0a-\u53f3\u4e0a-\u5de6\u4e0b-\u53f3\u4e0b”<\/td>\r\n\u6309\u7edf\u4e00\u89c4\u8303\u91cd\u65b0\u6392\u5e8f<\/td>\r\n<\/tr>\r\n
      \u5750\u6807\u8d8a\u754c<\/td>\r\n\u4e2a\u522b\u6807\u6ce8\u70b9\u8d85\u51fa\u56fe\u50cf\u8fb9\u754c<\/td>\r\n\u88c1\u526a\u81f3\u56fe\u50cf\u8303\u56f4\u5185<\/td>\r\n<\/tr>\r\n<\/tbody>\r\n<\/table>\r\n<\/figure>\r\n\r\n\r\n\r\n

      \u6570\u636e\u6e05\u6d17\u6d41\u7a0b\uff1a<\/strong><\/p>\r\n\r\n\r\n\r\n

        \r\n
      1. \u904d\u5386\u6240\u6709 .mat<\/code> \u6807\u7b7e\u6587\u4ef6\uff0c\u68c0\u67e5 points<\/code> \u6570\u7ec4\u7ef4\u5ea6\u662f\u5426\u4e3a [18, 4, 2]<\/code><\/li>\r\n
      2. \u6838\u5bf9\u56fe\u7247\u6587\u4ef6\u540d\u4e0e\u6807\u7b7e\u6587\u4ef6\u540d\u662f\u5426\u4e00\u4e00\u5bf9\u5e94<\/li>\r\n
      3. \u68c0\u67e5\u89d2\u70b9\u987a\u5e8f\u662f\u5426\u7b26\u5408”\u5de6\u4e0a-\u53f3\u4e0a-\u5de6\u4e0b-\u53f3\u4e0b”\u89c4\u8303<\/li>\r\n
      4. \u9a8c\u8bc1\u6240\u6709\u5750\u6807\u662f\u5426\u5728\u56fe\u50cf\u6709\u6548\u8303\u56f4\u5185<\/li>\r\n<\/ol>\r\n\r\n\r\n\r\n

        \u6700\u7ec8\u4fdd\u7559 983\u5f20\u56fe\u7247-\u6807\u7b7e\u5b8c\u5168\u5bf9\u5e94<\/strong> \u7684\u6570\u636e\u4f5c\u4e3a\u8bad\u7ec3\u96c6\u548c\u9a8c\u8bc1\u96c6\uff0c\u53e6\u6709\u90e8\u5206\u540c\u5b66\u4fdd\u7559\u7684\u539f\u59cb JSON \u683c\u5f0f\u6807\u6ce8\u6587\u4ef6\u4f5c\u4e3a\u72ec\u7acb\u6d4b\u8bd5\u96c6\u3002<\/p>\r\n\r\n\r\n\r\n

        \"\"<\/p>\r\n\r\n\r\n

        \r\n\r\n\r\n

        \u4e09\u3001\u6280\u672f\u6808<\/h2>\r\n\r\n\r\n\r\n
        \r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
        \u6280\u672f\/\u5de5\u5177<\/th>\r\n\u7248\u672c\/\u578b\u53f7<\/th>\r\n\u7528\u9014<\/th>\r\n<\/tr>\r\n<\/thead>\r\n
        YOLOv8n-pose<\/strong><\/td>\r\nUltralytics 8.x<\/td>\r\n\u810a\u67f118\u4e2a\u5173\u952e\u70b9\u68c0\u6d4b<\/td>\r\n<\/tr>\r\n
        SiliconFlow API<\/strong><\/td>\r\nQwen2.5-72B-Instruct<\/td>\r\n\u533b\u5b66\u62a5\u544a\u6587\u672c\u751f\u6210<\/td>\r\n<\/tr>\r\n
        Gradio<\/strong><\/td>\r\n6.x<\/td>\r\nWeb\u53ef\u89c6\u5316\u4ea4\u4e92\u754c\u9762<\/td>\r\n<\/tr>\r\n
        OpenCV<\/strong><\/td>\r\n4.x<\/td>\r\n\u56fe\u50cf\u5904\u7406\u4e0e\u53ef\u89c6\u5316\u6807\u6ce8<\/td>\r\n<\/tr>\r\n
        NumPy \/ SciPy<\/strong><\/td>\r\n\u2014<\/td>\r\n\u6570\u503c\u8ba1\u7b97\u4e0e\u810a\u67f1\u66f2\u7ebf\u62df\u5408<\/td>\r\n<\/tr>\r\n
        LabelMe<\/strong><\/td>\r\n\u2014<\/td>\r\n\u6570\u636e\u96c6\u6807\u6ce8<\/td>\r\n<\/tr>\r\n
        Python<\/strong><\/td>\r\n3.10<\/td>\r\n\u5f00\u53d1\u8bed\u8a00<\/td>\r\n<\/tr>\r\n<\/tbody>\r\n<\/table>\r\n<\/figure>\r\n\r\n\r\n
        \r\n\r\n\r\n

        \u56db\u3001\u6570\u636e\u9884\u5904\u7406\u4e0e\u683c\u5f0f\u8f6c\u6362<\/h2>\r\n\r\n\r\n\r\n

        4.1 \u4ece\u89d2\u70b9\u5230\u4e2d\u5fc3\u70b9<\/h3>\r\n\r\n\r\n\r\n

        YOLO-Pose \u6a21\u578b\u9700\u8981\u8f93\u5165\u5173\u952e\u70b9\u4e2d\u5fc3\u5750\u6807<\/strong>\u800c\u975e\u89d2\u70b9\uff0c\u56e0\u6b64\u9700\u8981\u5c06\u6bcf\u4e2a\u690e\u4f53\u76844\u4e2a\u89d2\u70b9\u8f6c\u6362\u4e3a1\u4e2a\u4e2d\u5fc3\u70b9\uff1a<\/p>\r\n\r\n\r\n\r\n

        # \u89d2\u70b9\u987a\u5e8f\uff1a\u5de6\u4e0a\u3001\u53f3\u4e0a\u3001\u5de6\u4e0b\u3001\u53f3\u4e0b\r\ncenter_x = (x1 + x2 + x3 + x4) \/ 4\r\ncenter_y = (y1 + y2 + y3 + y4) \/ 4<\/code><\/pre>\r\n\r\n\r\n\r\n
        4.2 \u8f6c\u6362\u4e3aYOLO\u683c\u5f0f<\/h3>\r\n\r\n\r\n\r\n
        YOLO-Pose \u8981\u6c42\u7684\u6807\u7b7e\u683c\u5f0f\u4e3a\uff1a<\/p>\r\n\r\n\r\n\r\n
        <class_id> <x1> <y1> <x2> <y2> ... <x18> <y18><\/code><\/pre>\r\n\r\n\r\n\r\n
        \u5176\u4e2d\u5750\u6807\u9700\u5f52\u4e00\u5316\u5230 [0, 1]<\/code> \u8303\u56f4\u3002\u7f16\u5199\u8f6c\u6362\u811a\u672c\u6279\u91cf\u5904\u7406983\u5f20\u6570\u636e\uff1a<\/p>\r\n\r\n\r\n\r\n
        import scipy.io as sio\r\nfrom pathlib import Path\r\n\r\ndef convert_mat_to_yolo(mat_path, img_w, img_h):\r\n    data = sio.loadmat(mat_path)\r\n    points = data['points']  # [18, 4, 2]\r\n\r\n    # \u8ba1\u7b97\u4e2d\u5fc3\u70b9\r\n    centers = points.mean(axis=1)  # [18, 2]\r\n\r\n    # \u5f52\u4e00\u5316\r\n    centers[:, 0] \/= img_w\r\n    centers[:, 1] \/= img_h\r\n\r\n    # \u6784\u5efaYOLO\u683c\u5f0f\u884c\r\n    yolo_line = \"0 \" + \" \".join([f\"{x:.6f} {y:.6f}\" for x, y in centers])\r\n    return yolo_line<\/code><\/pre>\r\n\r\n\r\n\r\n
        4.3 \u6570\u636e\u96c6\u5212\u5206<\/h3>\r\n\r\n\r\n\r\n
        \u6309\u7167 8:2<\/strong> \u7684\u6bd4\u4f8b\u5212\u5206\u8bad\u7ec3\u96c6\u4e0e\u9a8c\u8bc1\u96c6\uff1a<\/p>\r\n\r\n\r\n\r\n
        \r\n\r\n\r\n\r\n\r\n\r\n\r\n
        \u5b50\u96c6<\/th>\r\n\u6570\u91cf<\/th>\r\n\u6bd4\u4f8b<\/th>\r\n\u7528\u9014<\/th>\r\n<\/tr>\r\n<\/thead>\r\n
        \u8bad\u7ec3\u96c6<\/td>\r\n786\u5f20<\/td>\r\n80%<\/td>\r\n\u6a21\u578b\u8bad\u7ec3<\/td>\r\n<\/tr>\r\n
        \u9a8c\u8bc1\u96c6<\/td>\r\n197\u5f20<\/td>\r\n20%<\/td>\r\n\u8d85\u53c2\u6570\u8c03\u4f18\u4e0e\u8bad\u7ec3\u76d1\u63a7<\/td>\r\n<\/tr>\r\n<\/tbody>\r\n<\/table>\r\n<\/figure>\r\n\r\n\r\n\r\n
        \u6b64\u5916\uff0c\u53e6\u6709\u90e8\u5206\u540c\u5b66\u4fdd\u7559\u7684\u539f\u59cb JSON \u683c\u5f0f\u6807\u6ce8\u6587\u4ef6\u4f5c\u4e3a\u72ec\u7acb\u6d4b\u8bd5\u96c6\uff0c\u7528\u4e8e\u6700\u7ec8\u6548\u679c\u9a8c\u8bc1\u3002<\/p>\r\n\r\n\r\n\r\n
        \"\"<\/p>\r\n\r\n
        \u4e94\u3001\u6a21\u578b\u8bad\u7ec3<\/h2>\r\n\r\n\r\n\r\n
        5.1 \u6a21\u578b\u9009\u62e9\uff1aYOLOv8n-pose<\/h3>\r\n\r\n\r\n\r\n
        \u9009\u7528 YOLOv8n-pose<\/strong>\uff08nano\u7248\u672c\uff09\uff0c\u539f\u56e0\u5982\u4e0b\uff1a<\/p>\r\n\r\n\r\n\r\n
          \r\n
        • \u53c2\u6570\u91cf\u5c0f<\/strong>\uff08\u7ea63.2M\uff09\uff0c\u63a8\u7406\u901f\u5ea6\u5feb\uff0c\u9002\u5408\u90e8\u7f72<\/li>\r\n
        • \u5355\u9636\u6bb5\u68c0\u6d4b<\/strong>\uff1a\u7aef\u5230\u7aef\u8bad\u7ec3\uff0c\u65e0\u9700\u590d\u6742\u540e\u5904\u7406<\/li>\r\n
        • \u533b\u5b66\u5f71\u50cf\u9002\u914d<\/strong>\uff1a\u5bf9X\u5149\u7247\u8fd9\u7c7b\u7070\u5ea6\u56fe\u50cf\u7684\u5173\u952e\u70b9\u68c0\u6d4b\u6548\u679c\u826f\u597d<\/li>\r\n<\/ul>\r\n\r\n\r\n\r\n
          5.2 \u8bad\u7ec3\u914d\u7f6e<\/h3>\r\n\r\n\r\n\r\n
          \u521b\u5efa data.yaml<\/code> \u914d\u7f6e\u6587\u4ef6\uff1a<\/p>\r\n\r\n\r\n\r\n
          path: C:\\python_ws\\\u8ba1\u7b97\u673a\u89c6\u89c9\u5b9e\u8bad\\SpineVLM\\dataset\r\ntrain: train_yolo\/images\r\nval: val_yolo\/images\r\ntest: test_yolo\/images\r\n\r\nkpt_shape: [18, 2]  # 18\u4e2a\u5173\u952e\u70b9\uff0c2D\u5750\u6807\r\nflip_idx: [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17]\r\n\r\nnames:\r\n  0: spine<\/code><\/pre>\r\n\r\n\r\n\r\n
          \u5173\u952e\u8bad\u7ec3\u53c2\u6570\uff1a<\/p>\r\n\r\n\r\n\r\n
            \r\n
          • \u8f93\u5165\u5c3a\u5bf8\uff1a640\u00d7640<\/li>\r\n
          • \u6279\u6b21\u5927\u5c0f\uff1a16<\/li>\r\n
          • \u8bad\u7ec3\u8f6e\u6570\uff1a300 epoch<\/li>\r\n
          • \u4f18\u5316\u5668\uff1aSGD\uff0c\u521d\u59cb\u5b66\u4e60\u7387 0.01<\/li>\r\n
          • \u6570\u636e\u589e\u5f3a\uff1aMosaic\u3001MixUp\u3001\u968f\u673a\u6c34\u5e73\u7ffb\u8f6c<\/li>\r\n<\/ul>\r\n\r\n\r\n\r\n
            5.3 \u8bad\u7ec3\u7ed3\u679c<\/h3>\r\n\r\n\r\n\r\n
            \r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
            \u6307\u6807<\/th>\r\n\u6570\u503c<\/th>\r\n<\/tr>\r\n<\/thead>\r\n
            mAP@0.5<\/td>\r\n0.92<\/td>\r\n<\/tr>\r\n
            mAP@0.5:0.95<\/td>\r\n0.78<\/td>\r\n<\/tr>\r\n
            \u5173\u952e\u70b9mAP<\/td>\r\n0.85<\/td>\r\n<\/tr>\r\n
            \u63a8\u7406\u901f\u5ea6<\/td>\r\n~45ms\/\u5f20\uff08GPU\uff09<\/td>\r\n<\/tr>\r\n<\/tbody>\r\n<\/table>\r\n<\/figure>\r\n\r\n\r\n\r\n
            \u6a21\u578b\u80fd\u591f\u7a33\u5b9a\u68c0\u6d4b18\u4e2a\u810a\u67f1\u5173\u952e\u70b9\uff0c\u5b9a\u4f4d\u7cbe\u5ea6\u6ee1\u8db3\u540e\u7eed\u4e34\u5e8a\u53c2\u6570\u8ba1\u7b97\u9700\u6c42\u3002<\/p>\r\n
            \"\"<\/p>\r\n\r\n\r\n\r\n
             <\/p>\r\n\r\n\r\n\r\n
            \"\"<\/em><\/p>\r\n\r\n\r\n
            \r\n\r\n\r\n
            \u516d\u3001\u4e34\u5e8a\u53c2\u6570\u8ba1\u7b97\uff1a\u4ece\u5173\u952e\u70b9\u5230\u8bca\u65ad\u6307\u6807<\/h2>\r\n\r\n\r\n\r\n
            \u6a21\u578b\u8f93\u51fa\u7684\u662f18\u4e2a\u690e\u4f53\u7684\u4e2d\u5fc3\u70b9\u5750\u6807\uff0c\u9700\u8981\u8fdb\u4e00\u6b65\u8ba1\u7b97\u4e34\u5e8a\u53c2\u6570\u3002\u4ee5\u4e0b\u662f\u6838\u5fc3\u53c2\u6570\u7684\u8ba1\u7b97\u65b9\u6cd5\uff1a<\/p>\r\n\r\n\r\n\r\n
            6.1 Cobb\u89d2\uff08\u810a\u67f1\u4fa7\u5f2f\u91d1\u6807\u51c6\uff09<\/h3>\r\n\r\n\r\n\r\n
            \u80f8Cobb\u89d2<\/strong> = T5\u4e0a\u7ec8\u677f\u4e0eT12\u4e0b\u7ec8\u677f\u7684\u5939\u89d2 \u8170Cobb\u89d2<\/strong> = T12\u4e0a\u7ec8\u677f\u4e0eL4\u4e0b\u7ec8\u677f\u7684\u5939\u89d2<\/p>\r\n\r\n\r\n\r\n
            \u5b9e\u73b0\u4e2d\uff0c\u7ec8\u677f\u65b9\u5411\u901a\u8fc7\u76f8\u90bb\u690e\u4f53\u4e2d\u5fc3\u70b9\u8fde\u7ebf\u7684\u5782\u7ebf\u786e\u5b9a\uff1a<\/p>\r\n\r\n\r\n\r\n
            def vector(p1, p2):\r\n    return np.array([p2[0] - p1[0], p2[1] - p1[1]])\r\n\r\ndef perpendicular(v):\r\n    \"\"\"\u5411\u91cf\u65cb\u8f6c90\u5ea6\u5f97\u5230\u7ec8\u677f\u7ebf\u65b9\u5411\"\"\"\r\n    return np.array([-v[1], v[0]])\r\n\r\n# T5\u4e0a\u7ec8\u677f\u65b9\u5411\uff08\u5782\u76f4\u4e8eT4-T5\u8fde\u7ebf\uff09\r\nt5_long = vector(kp[T4], kp[T5])\r\nt5_upper = perpendicular(t5_long)\r\n\r\n# T12\u4e0b\u7ec8\u677f\u65b9\u5411\uff08\u5782\u76f4\u4e8eT12-L1\u8fde\u7ebf\uff09\r\nt12_long = vector(kp[T12], kp[L1])\r\nt12_lower = perpendicular(t12_long)\r\n\r\nthoracic_cobb = angle_between(t5_upper, t12_lower)<\/code><\/pre>\r\n\r\n\r\n\r\n
            \u4e34\u5e8a\u53c2\u8003\u503c\uff1a<10\u00b0<\/code> \u6b63\u5e38\uff0c10-25\u00b0<\/code> \u8f7b\u5ea6\uff0c25-40\u00b0<\/code> \u4e2d\u5ea6\uff0c>40\u00b0<\/code> \u91cd\u5ea6<\/p>\r\n\r\n\r\n\r\n
            6.2 C7PL-CSVL\uff08\u51a0\u72b6\u9762\u5e73\u8861\uff09<\/h3>\r\n\r\n\r\n\r\n
            C7\u94c5\u5782\u7ebf\u4e0e\u9ab6\u9aa8\u4e2d\u5782\u7ebf\uff08CSVL\uff09\u7684\u6c34\u5e73\u504f\u79fb\u8ddd\u79bb\uff1a<\/p>\r\n\r\n\r\n\r\n
            c7pl_csvl_px = abs(c7_center[0] - l5_center[0])\r\nc7pl_csvl_mm = c7pl_csvl_px * 0.3  # \u50cf\u7d20\u5230\u6beb\u7c73\u8f6c\u6362<\/code><\/pre>\r\n\r\n\r\n\r\n
            \u4e34\u5e8a\u53c2\u8003\u503c\uff1a<20mm<\/code> \u6b63\u5e38\uff0c20-40mm<\/code> \u8f7b\u5ea6\u5931\u8861\uff0c>40mm<\/code> \u660e\u663e\u5931\u8861<\/p>\r\n\r\n\r\n\r\n
            6.3 T1\u503e\u659c\u89d2<\/h3>\r\n\r\n\r\n\r\n
            T1\u4e0a\u7ec8\u677f\u4e0e\u6c34\u5e73\u7ebf\u7684\u5939\u89d2\uff0c\u53cd\u6620\u4e0a\u80f8\u690e\u6bb5\u503e\u659c\u7a0b\u5ea6\uff1a<\/p>\r\n\r\n\r\n\r\n
            t1_long = vector(kp[C7], kp[T1])\r\nt1_upper = perpendicular(t1_long)\r\nt1_tilt = line_angle_with_horizontal(t1_upper)<\/code><\/pre>\r\n\r\n\r\n\r\n
            6.4 \u810a\u67f1\u66f2\u7ebf\u62df\u5408<\/h3>\r\n\r\n\r\n\r\n
            \u4f7f\u7528 UnivariateSpline<\/strong> \u5bf918\u4e2a\u5173\u952e\u70b9\u8fdb\u884c\u4e09\u6b21\u6837\u6761\u62df\u5408\uff0c\u8ba1\u7b97\u66f2\u7387\u5206\u5e03\uff0c\u8bc6\u522b\u6700\u5927\u5f2f\u66f2\u4f4d\u7f6e\uff1a<\/p>\r\n\r\n\r\n\r\n
            from scipy.interpolate import UnivariateSpline\r\nfrom scipy.signal import find_peaks\r\n\r\nspline = UnivariateSpline(y_sorted, x_sorted, k=3, s=10)\r\ncurvature = np.abs(d2x_dy2) \/ (1 + dx_dy**2)**1.5\r\npeaks, _ = find_peaks(curvature, distance=50)<\/code><\/pre>\r\n\r\n\r\n\r\n\r\n
            \r\n\r\n\r\n
            \u4e03\u3001\u53ef\u89c6\u5316\u7cfb\u7edf\uff1a\u591a\u8f6e\u8fed\u4ee3\u6253\u78e8\u7ec6\u8282<\/h2>\r\n\r\n\r\n\r\n
            \u53ef\u89c6\u5316\u90e8\u5206\u7ecf\u5386\u4e86\u591a\u8f6e\u8fed\u4ee3\u4f18\u5316\uff0c\u6bcf\u4e00\u8f6e\u90fd\u662f\u57fa\u4e8e\u5b9e\u9645\u4f7f\u7528\u53cd\u9988\u8fdb\u884c\u8c03\u6574\uff1a<\/p>\r\n\r\n\r\n\r\n
            \r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
            \u8fed\u4ee3\u7248\u672c<\/th>\r\n\u95ee\u9898\u53cd\u9988<\/th>\r\n\u89e3\u51b3\u65b9\u6848<\/th>\r\n<\/tr>\r\n<\/thead>\r\n
            v1<\/td>\r\n\u5927\u65b9\u6846\u6846\u4f4f\u6574\u4e2a\u810a\u67f1\uff0c\u770b\u4e0d\u6e05\u5355\u4e2a\u690e\u4f53<\/td>\r\n\u6539\u4e3a\u6bcf\u4e2a\u690e\u4f53\u72ec\u7acb\u56db\u8fb9\u5f62\u6846<\/td>\r\n<\/tr>\r\n
            v2<\/td>\r\n\u53c2\u8003\u7ebf\u4e0e\u690e\u4f53\u6846\u91cd\u5408\uff0c\u53c2\u6570\u503c\u770b\u4e0d\u6e05<\/td>\r\n\u5ef6\u957f\u7ec8\u677f\u7ebf\uff08extend_factor=2.5\uff09\uff0c\u589e\u52a0\u767d\u8272\u63cf\u8fb9<\/td>\r\n<\/tr>\r\n
            v3<\/td>\r\n\u53c2\u6570\u9762\u677f\u88ab\u622a\u65ad\uff0c\u9ed1\u5e95\u9ed1\u5b57\u770b\u4e0d\u6e05<\/td>\r\n\u6539\u4e3a\u767d\u5e95\u9ed1\u5b57\uff0c\u79fb\u81f3\u5de6\u4e0b\u89d2<\/td>\r\n<\/tr>\r\n
            v4<\/td>\r\n\u5173\u952e\u70b9\u6807\u8bb0\u50cf”\u9ed1\u5708\u9ed1\u5341\u5b57”<\/td>\r\n\u6539\u4e3a\u9ec4\u5708\u767d\u5341\u5b57\u9ed1\u7ec6\u7ebf<\/td>\r\n<\/tr>\r\n
            v5<\/td>\r\n\u539f\u56fe\u88ab\u8986\u76d6\u770b\u4e0d\u6e05<\/td>\r\n\u964d\u4f4e\u6240\u6709\u53e0\u52a0\u5143\u7d20\u900f\u660e\u5ea6\uff0c\u4ec5\u753b\u7ec6\u7ebf<\/td>\r\n<\/tr>\r\n
            v6<\/td>\r\n\u53c2\u6570\u9762\u677f\u906e\u6321\u56fe\u50cf<\/td>\r\n\u53c2\u6570\u76f4\u63a5\u6807\u5728\u5bf9\u5e94\u4f4d\u7f6e\uff0c\u53d6\u6d88\u72ec\u7acb\u9762\u677f<\/td>\r\n<\/tr>\r\n<\/tbody>\r\n<\/table>\r\n<\/figure>\r\n\r\n\r\n\r\n
            \u6700\u7ec8\u53ef\u89c6\u5316\u6548\u679c\u5305\u542b\u4ee5\u4e0b\u5143\u7d20\uff1a<\/strong><\/p>\r\n\r\n\r\n\r\n