{"id":35,"date":"2025-07-31T08:05:07","date_gmt":"2025-07-31T16:05:07","guid":{"rendered":"https:\/\/dreamrunrun.com\/blog\/?p=35"},"modified":"2025-08-05T06:01:55","modified_gmt":"2025-08-05T14:01:55","slug":"%e5%9f%ba%e4%ba%8eopencv%e5%ae%9e%e7%8e%b0%e4%ba%ba%e8%84%b8%e7%89%b9%e5%be%81%e8%af%86%e5%88%ab","status":"publish","type":"post","link":"https:\/\/dreamrunrun.com\/blog\/?p=35","title":{"rendered":"\u57fa\u4e8eopencv\u5b9e\u73b0\u4eba\u8138\u7279\u5f81\u8bc6\u522b"},"content":{"rendered":"\n<p class=\"wp-block-paragraph\">\u8fd9\u91cc\u6211\u4eec\u4e3b\u8981\u5173\u6ce8\u4e24\u79cd\u6700\u5e38\u89c1\u7684\u201c\u7279\u5f81\u8bc6\u522b\u201d\uff1a<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li><strong>\u4eba\u8138\u68c0\u6d4b<\/strong>\uff1a\u5728\u56fe\u50cf\u6216\u89c6\u9891\u4e2d\u627e\u5230\u4eba\u8138\u7684\u4f4d\u7f6e\u548c\u5927\u5c0f\uff0c\u5e76\u7528\u4e00\u4e2a\u65b9\u6846\u6846\u51fa\u6765\u3002\u8fd9\u662f\u6240\u6709\u540e\u7eed\u64cd\u4f5c\u7684\u57fa\u7840\u3002<\/li>\n\n\n\n<li><strong>\u4eba\u8138\u5173\u952e\u70b9\u68c0\u6d4b<\/strong>\uff1a\u5728\u68c0\u6d4b\u5230\u7684\u4eba\u8138\u4e0a\uff0c\u5b9a\u4f4d\u51fa\u5173\u952e\u7684\u7279\u5f81\u70b9\uff0c\u5982\u773c\u775b\u3001\u9f3b\u5b50\u3001\u5634\u5df4\u3001\u7709\u6bdb\u7684\u8f6e\u5ed3\u7b49\u3002\u8fd9\u901a\u5e38\u88ab\u79f0\u4e3a\u201c\u4eba\u8138\u5bf9\u9f50\u201d\u6216\u201c\u7279\u5f81\u70b9\u6807\u8bb0\u201d\u3002<\/li>\n<\/ol>\n\n\n\n<p class=\"wp-block-paragraph\">\u6211\u4eec\u5c06\u4ece\u6700\u57fa\u7840\u3001\u6700\u7ecf\u5178\u7684\u65b9\u6cd5\u5f00\u59cb\uff0c\u9010\u6b65\u8fc7\u6e21\u5230\u66f4\u73b0\u4ee3\u3001\u66f4\u7cbe\u786e\u7684\u6df1\u5ea6\u5b66\u4e60\u65b9\u6cd5\u3002<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h3 class=\"wp-block-heading\">\u51c6\u5907\u5de5\u4f5c\uff1a\u5b89\u88c5 OpenCV<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">\u9996\u5148\uff0c\u786e\u4fdd\u4f60\u5df2\u7ecf\u5b89\u88c5\u4e86 OpenCV \u7684 Python \u5e93\u3002\u5982\u679c\u5c1a\u672a\u5b89\u88c5\uff0c\u53ef\u4ee5\u901a\u8fc7 pip \u8fdb\u884c\u5b89\u88c5\u3002\u63a8\u8350\u5b89\u88c5\u5305\u542b\u989d\u5916\u6a21\u5757\u7684&nbsp;<code>opencv-contrib-python<\/code>\uff0c\u56e0\u4e3a\u5b83\u5305\u542b\u4e86\u66f4\u591a\u7684\u4eba\u8138\u8bc6\u522b\u76f8\u5173\u7b97\u6cd5\u3002<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>pip install opencv-contrib-python\n<\/code><\/pre>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h3 class=\"wp-block-heading\">\u65b9\u6cd5\u4e00\uff1a\u7ecf\u5178\u7684\u4eba\u8138\u68c0\u6d4b (Haar \u7ea7\u8054\u5206\u7c7b\u5668)<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">\u8fd9\u662f OpenCV \u4e2d\u6700\u4f20\u7edf\u3001\u6700\u7ecf\u5178\u7684\u4eba\u8138\u68c0\u6d4b\u65b9\u6cd5\u3002\u5b83\u57fa\u4e8e Haar \u7279\u5f81\u548c AdaBoost \u7b97\u6cd5\uff0c\u901a\u8fc7\u8bad\u7ec3\u4e00\u4e2a\u7ea7\u8054\u5206\u7c7b\u5668\u6765\u5feb\u901f\u68c0\u6d4b\u4eba\u8138\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>\u4f18\u70b9<\/strong>\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u901f\u5ea6\u5feb\uff0c\u5728 CPU \u4e0a\u5c31\u80fd\u5b9e\u65f6\u8fd0\u884c\u3002<\/li>\n\n\n\n<li>\u65e0\u9700\u989d\u5916\u4f9d\u8d56\uff0c\u6a21\u578b\u6587\u4ef6\u5305\u542b\u5728 OpenCV \u4e2d\u3002<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>\u7f3a\u70b9<\/strong>\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u7cbe\u5ea6\u76f8\u5bf9\u8f83\u4f4e\uff0c\u5bb9\u6613\u53d7\u5149\u7167\u3001\u59ff\u6001\u3001\u906e\u6321\u5f71\u54cd\u3002<\/li>\n\n\n\n<li>\u53ea\u80fd\u68c0\u6d4b\u5230\u4eba\u8138\u7684\u77e9\u5f62\u6846\uff0c\u65e0\u6cd5\u63d0\u4f9b\u5173\u952e\u70b9\u4fe1\u606f\u3002<\/li>\n<\/ul>\n\n\n\n<h4 class=\"wp-block-heading\">\u5b9e\u73b0\u6b65\u9aa4<\/h4>\n\n\n\n<ol class=\"wp-block-list\">\n<li><strong>\u52a0\u8f7d\u9884\u8bad\u7ec3\u6a21\u578b<\/strong>\uff1aOpenCV \u63d0\u4f9b\u4e86\u8bad\u7ec3\u597d\u7684 Haar \u7ea7\u8054\u5206\u7c7b\u5668 XML \u6587\u4ef6\u3002<\/li>\n\n\n\n<li><strong>\u8bfb\u53d6\u56fe\u50cf<\/strong>\uff1a\u4f7f\u7528&nbsp;<code>cv2.imread()<\/code>&nbsp;\u52a0\u8f7d\u4e00\u5f20\u56fe\u7247\u3002<\/li>\n\n\n\n<li><strong>\u7070\u5ea6\u5316\u5904\u7406<\/strong>\uff1aHaar \u5206\u7c7b\u5668\u5728\u7070\u5ea6\u56fe\u50cf\u4e0a\u8fd0\u884c\uff0c\u9700\u8981\u5c06\u5f69\u8272\u56fe\u50cf\u8f6c\u6362\u4e3a\u7070\u5ea6\u56fe\u3002<\/li>\n\n\n\n<li><strong>\u6267\u884c\u68c0\u6d4b<\/strong>\uff1a\u4f7f\u7528&nbsp;<code>cv2.CascadeClassifier.detectMultiScale()<\/code>&nbsp;\u65b9\u6cd5\u8fdb\u884c\u4eba\u8138\u68c0\u6d4b\u3002<\/li>\n\n\n\n<li><strong>\u7ed8\u5236\u7ed3\u679c<\/strong>\uff1a\u5728\u68c0\u6d4b\u5230\u7684\u4eba\u8138\u5468\u56f4\u7ed8\u5236\u77e9\u5f62\u6846\u3002<\/li>\n<\/ol>\n\n\n\n<h4 class=\"wp-block-heading\">\u4ee3\u7801\u793a\u4f8b<\/h4>\n\n\n\n<pre class=\"wp-block-code\"><code>import cv2\n\n<em># 1. \u52a0\u8f7d\u9884\u8bad\u7ec3\u7684 Haar \u7ea7\u8054\u5206\u7c7b\u5668\u6a21\u578b<\/em>\n<em># 'haarcascade_frontalface_default.xml' \u662f OpenCV \u81ea\u5e26\u7684\u7528\u4e8e\u68c0\u6d4b\u6b63\u8138\u7684\u6a21\u578b<\/em>\n<em># \u4f60\u53ef\u4ee5\u5728 opencv-python \u7684\u5b89\u88c5\u76ee\u5f55\u4e0b\u627e\u5230\u5b83\uff0c\u6216\u8005\u4ece\u7f51\u4e0a\u4e0b\u8f7d<\/em>\nface_cascade_path = cv2.data.haarcascades + 'haarcascade_frontalface_default.xml'\nface_cascade = cv2.CascadeClassifier(face_cascade_path)\n\n<em># 2. \u8bfb\u53d6\u56fe\u50cf<\/em>\n<em># \u8bf7\u5c06 'your_image.jpg' \u66ff\u6362\u4e3a\u4f60\u7684\u56fe\u7247\u8def\u5f84<\/em>\nimage_path = 'your_image.jpg'\nimage = cv2.imread(image_path)\n\nif image is None:\n    print(f\"\u9519\u8bef\uff1a\u65e0\u6cd5\u5728\u8def\u5f84 '{image_path}' \u627e\u5230\u56fe\u50cf\u3002\")\n    exit()\n\n<em># 3. \u5c06\u56fe\u50cf\u8f6c\u6362\u4e3a\u7070\u5ea6\u56fe<\/em>\ngray_image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)\n\n<em># 4. \u6267\u884c\u4eba\u8138\u68c0\u6d4b<\/em>\n<em># detectMultiScale \u53c2\u6570\u8bf4\u660e\uff1a<\/em>\n<em># gray_image: \u8f93\u5165\u7684\u7070\u5ea6\u56fe\u50cf<\/em>\n<em># scaleFactor: \u6bcf\u6b21\u56fe\u50cf\u5c3a\u5bf8\u51cf\u5c0f\u7684\u6bd4\u4f8b\uff0c\u7528\u4e8e\u6784\u5efa\u56fe\u50cf\u91d1\u5b57\u5854<\/em>\n<em># minNeighbors: \u6bcf\u4e2a\u5019\u9009\u77e9\u5f62\u5e94\u5305\u542b\u7684\u90bb\u8fd1\u5019\u9009\u6846\u4e2a\u6570\uff0c\u7528\u4e8e\u6291\u5236\u5f31\u68c0\u6d4b<\/em>\n<em># minSize: \u53ef\u80fd\u7684\u6700\u5c0f\u4eba\u8138\u5c3a\u5bf8<\/em>\nfaces = face_cascade.detectMultiScale(\n    gray_image,\n    scaleFactor=1.1,\n    minNeighbors=5,\n    minSize=(30, 30)\n)\n\nprint(f\"\u68c0\u6d4b\u5230 {len(faces)} \u5f20\u4eba\u8138\")\n\n<em># 5. \u5728\u539f\u59cb\u56fe\u50cf\u4e0a\u7ed8\u5236\u68c0\u6d4b\u7ed3\u679c<\/em>\nfor (x, y, w, h) in faces:\n    <em># cv2.rectangle(\u56fe\u50cf, \u5de6\u4e0a\u89d2\u5750\u6807, \u53f3\u4e0b\u89d2\u5750\u6807, \u989c\u8272, \u7ebf\u6761\u7c97\u7ec6)<\/em>\n    cv2.rectangle(image, (x, y), (x+w, y+h), (0, 255, 0), 2)\n\n<em># \u663e\u793a\u7ed3\u679c\u56fe\u50cf<\/em>\ncv2.imshow('Haar Face Detection', image)\n\n<em># \u7b49\u5f85\u6309\u952e\uff0c\u7136\u540e\u5173\u95ed\u6240\u6709\u7a97\u53e3<\/em>\ncv2.waitKey(0)\ncv2.destroyAllWindows()\n<\/code><\/pre>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h3 class=\"wp-block-heading\">\u65b9\u6cd5\u4e8c\uff1a\u4eba\u8138\u5173\u952e\u70b9\u68c0\u6d4b (Dlib \u5e93)<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">\u867d\u7136 OpenCV \u672c\u8eab\u4e5f\u6709\u5173\u952e\u70b9\u68c0\u6d4b\u6a21\u578b\uff08\u5982 LBF\uff09\uff0c\u4f46\u4e1a\u754c\u66f4\u5e38\u7528\u3001\u6548\u679c\u66f4\u597d\u7684\u662f&nbsp;<strong>Dlib<\/strong>&nbsp;\u5e93\u3002\u5b83\u63d0\u4f9b\u4e86\u975e\u5e38\u7cbe\u786e\u7684 68 \u70b9\u4eba\u8138\u5173\u952e\u70b9\u68c0\u6d4b\u5668\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>\u4f18\u70b9<\/strong>\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u7cbe\u5ea6\u6781\u9ad8\uff0c\u5bf9\u59ff\u6001\u548c\u5149\u7167\u53d8\u5316\u6709\u8f83\u597d\u7684\u9c81\u68d2\u6027\u3002<\/li>\n\n\n\n<li>\u63d0\u4f9b\u4e86\u6807\u51c6\u5316\u7684 68 \u4e2a\u5173\u952e\u70b9\uff0c\u8986\u76d6\u4e86\u7709\u6bdb\u3001\u773c\u775b\u3001\u9f3b\u5b50\u3001\u5634\u5df4\u548c\u4e0b\u5df4\u7684\u8f6e\u5ed3\u3002<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>\u7f3a\u70b9<\/strong>\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u901f\u5ea6\u6bd4 Haar \u6162\uff0c\u5c24\u5176\u662f\u5728\u6ca1\u6709 GPU \u52a0\u901f\u7684\u60c5\u51b5\u4e0b\u3002<\/li>\n\n\n\n<li>\u9700\u8981\u989d\u5916\u5b89\u88c5 Dlib \u5e93\u3002<\/li>\n<\/ul>\n\n\n\n<h4 class=\"wp-block-heading\">\u51c6\u5907\u5de5\u4f5c\uff1a\u5b89\u88c5 Dlib<\/h4>\n\n\n\n<p class=\"wp-block-paragraph\">Dlib \u7684\u5b89\u88c5\u6709\u65f6\u4f1a\u6bd4\u8f83\u590d\u6742\uff0c\u5c24\u5176\u662f\u5728 Windows \u4e0a\u3002<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code><em># \u5bf9\u4e8e macOS \u548c Linux\uff0c\u901a\u5e38\u53ef\u4ee5\u76f4\u63a5\u901a\u8fc7 pip \u5b89\u88c5<\/em>\npip install dlib\n\n<em># \u5bf9\u4e8e Windows\uff0c\u5982\u679c\u76f4\u63a5\u5b89\u88c5\u5931\u8d25\uff0c\u4f60\u53ef\u80fd\u9700\u8981\uff1a<\/em>\n<em># 1. \u5b89\u88c5 CMake (https:\/\/cmake.org\/download\/)<\/em>\n<em># 2. \u5b89\u88c5 Visual Studio Build Tools<\/em>\n<em># 3. \u7136\u540e\u518d\u5c1d\u8bd5 pip install dlib<\/em>\n<em># \u6216\u8005\uff0c\u53ef\u4ee5\u5bfb\u627e\u9884\u7f16\u8bd1\u7684 whl \u6587\u4ef6\u8fdb\u884c\u5b89\u88c5\u3002<\/em>\n<\/code><\/pre>\n\n\n\n<p class=\"wp-block-paragraph\">\u4f60\u8fd8\u9700\u8981\u4e0b\u8f7d Dlib \u7684\u9884\u8bad\u7ec3\u6a21\u578b\u6587\u4ef6\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong><a href=\"http:\/\/dlib.net\/files\/shape_predictor_68_face_landmarks.dat.bz2\">shape_predictor_68_face_landmarks.dat<\/a><\/strong>&nbsp;(\u4e0b\u8f7d\u540e\u9700\u8981\u89e3\u538b)<\/li>\n<\/ul>\n\n\n\n<h4 class=\"wp-block-heading\">\u5b9e\u73b0\u6b65\u9aa4<\/h4>\n\n\n\n<ol class=\"wp-block-list\">\n<li><strong>\u52a0\u8f7d\u6a21\u578b<\/strong>\uff1a\u52a0\u8f7d Dlib \u7684\u4eba\u8138\u68c0\u6d4b\u5668\u548c\u5173\u952e\u70b9\u9884\u6d4b\u5668\u3002<\/li>\n\n\n\n<li><strong>\u68c0\u6d4b\u4eba\u8138<\/strong>\uff1a\u4f7f\u7528 Dlib \u7684\u68c0\u6d4b\u5668\u627e\u5230\u4eba\u8138\u4f4d\u7f6e\u3002<\/li>\n\n\n\n<li><strong>\u9884\u6d4b\u5173\u952e\u70b9<\/strong>\uff1a\u5bf9\u4e8e\u6bcf\u4e2a\u68c0\u6d4b\u5230\u7684\u4eba\u8138\uff0c\u4f7f\u7528\u5173\u952e\u70b9\u9884\u6d4b\u5668\u83b7\u53d6 68 \u4e2a\u70b9\u7684\u5750\u6807\u3002<\/li>\n\n\n\n<li><strong>\u7ed8\u5236\u5173\u952e\u70b9<\/strong>\uff1a\u5728\u56fe\u50cf\u4e0a\u5c06\u8fd9\u4e9b\u70b9\u7ed8\u5236\u51fa\u6765\u3002<\/li>\n<\/ol>\n\n\n\n<h4 class=\"wp-block-heading\">\u4ee3\u7801\u793a\u4f8b<\/h4>\n\n\n\n<pre class=\"wp-block-code\"><code>import cv2\nimport dlib\n\n<em># 1. \u52a0\u8f7d Dlib \u7684\u9884\u8bad\u7ec3\u6a21\u578b<\/em>\n<em># \u4f60\u9700\u8981\u5148\u4e0b\u8f7d 'shape_predictor_68_face_landmarks.dat' \u6587\u4ef6<\/em>\npredictor_path = 'shape_predictor_68_face_landmarks.dat'\n\n<em># \u521b\u5efa Dlib \u7684\u4eba\u8138\u68c0\u6d4b\u5668 (\u57fa\u4e8e HOG)<\/em>\ndetector = dlib.get_frontal_face_detector()\n<em># \u521b\u5efa\u5173\u952e\u70b9\u9884\u6d4b\u5668<\/em>\npredictor = dlib.shape_predictor(predictor_path)\n\n<em># 2. \u8bfb\u53d6\u56fe\u50cf<\/em>\nimage_path = 'your_image.jpg'\nimage = cv2.imread(image_path)\n\nif image is None:\n    print(f\"\u9519\u8bef\uff1a\u65e0\u6cd5\u5728\u8def\u5f84 '{image_path}' \u627e\u5230\u56fe\u50cf\u3002\")\n    exit()\n\n<em># Dlib \u5728\u7070\u5ea6\u56fe\u50cf\u4e0a\u5de5\u4f5c<\/em>\ngray_image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)\n\n<em># 3. \u68c0\u6d4b\u4eba\u8138<\/em>\n<em># detector \u7684\u7b2c\u4e8c\u4e2a\u53c2\u6570\u662f\u4e0a\u91c7\u6837\u6b21\u6570\uff0c\u53ef\u4ee5\u63d0\u9ad8\u68c0\u6d4b\u5c0f\u8138\u7684\u80fd\u529b\uff0c\u4f46\u4f1a\u53d8\u6162<\/em>\nfaces = detector(gray_image, 1)\n\nprint(f\"\u68c0\u6d4b\u5230 {len(faces)} \u5f20\u4eba\u8138\")\n\n<em># 4. \u904d\u5386\u68c0\u6d4b\u5230\u7684\u6bcf\u5f20\u4eba\u8138\uff0c\u5e76\u9884\u6d4b\u5173\u952e\u70b9<\/em>\nfor face in faces:\n    <em># \u9884\u6d4b 68 \u4e2a\u5173\u952e\u70b9<\/em>\n    landmarks = predictor(gray_image, face)\n    \n    <em># 5. \u7ed8\u5236\u5173\u952e\u70b9<\/em>\n    <em># landmarks.parts() \u662f\u4e00\u4e2a\u5305\u542b 68 \u4e2a\u70b9\u7684\u5bf9\u8c61<\/em>\n    for n in range(68):\n        <em># \u83b7\u53d6\u7b2c n \u4e2a\u70b9\u7684 (x, y) \u5750\u6807<\/em>\n        x = landmarks.part(n).x\n        y = landmarks.part(n).y\n        <em># \u5728\u56fe\u50cf\u4e0a\u753b\u4e00\u4e2a\u5b9e\u5fc3\u5706\u70b9<\/em>\n        cv2.circle(image, (x, y), 2, (0, 255, 0), -1)\n\n<em># \u663e\u793a\u7ed3\u679c\u56fe\u50cf<\/em>\ncv2.imshow('Dlib 68-Point Landmarks', image)\n\n<em># \u7b49\u5f85\u6309\u952e\uff0c\u7136\u540e\u5173\u95ed\u6240\u6709\u7a97\u53e3<\/em>\ncv2.waitKey(0)\ncv2.destroyAllWindows()\n<\/code><\/pre>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h3 class=\"wp-block-heading\">\u65b9\u6cd5\u4e09\uff1a\u73b0\u4ee3\u7684\u6df1\u5ea6\u5b66\u4e60\u65b9\u6cd5 (OpenCV DNN \u6a21\u5757)<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">\u8fd9\u662f\u76ee\u524d\u6700\u63a8\u8350\u7684\u65b9\u6cd5\uff0c\u5b83\u5728\u7cbe\u5ea6\u548c\u901f\u5ea6\u4e4b\u95f4\u53d6\u5f97\u4e86\u5f88\u597d\u7684\u5e73\u8861\u3002OpenCV \u7684 DNN (Deep Neural Network) \u6a21\u5757\u53ef\u4ee5\u52a0\u8f7d\u548c\u8fd0\u884c\u4e3b\u6d41\u6df1\u5ea6\u5b66\u4e60\u6846\u67b6\uff08\u5982 Caffe, TensorFlow, PyTorch\uff09\u8bad\u7ec3\u597d\u7684\u6a21\u578b\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u6211\u4eec\u5c06\u4f7f\u7528\u4e00\u4e2a\u57fa\u4e8e&nbsp;<strong>SSD (Single Shot MultiBox Detector)<\/strong>&nbsp;\u548c&nbsp;<strong>ResNet10<\/strong>&nbsp;\u9aa8\u5e72\u7f51\u7edc\u7684\u8f7b\u91cf\u7ea7\u4eba\u8138\u68c0\u6d4b\u6a21\u578b\uff0c\u5b83\u975e\u5e38\u5feb\u4e14\u51c6\u786e\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>\u4f18\u70b9<\/strong>\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u7cbe\u5ea6\u8fdc\u9ad8\u4e8e Haar \u7ea7\u8054\u5206\u7c7b\u5668\u3002<\/li>\n\n\n\n<li>\u901f\u5ea6\u975e\u5e38\u5feb\uff0c\u5728 CPU \u4e0a\u4e5f\u80fd\u8fbe\u5230\u5b9e\u65f6\u3002<\/li>\n\n\n\n<li>\u5bf9\u59ff\u6001\u3001\u5149\u7167\u3001\u906e\u6321\u7684\u9c81\u68d2\u6027\u66f4\u597d\u3002<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>\u7f3a\u70b9<\/strong>\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u9700\u8981\u4e0b\u8f7d\u6a21\u578b\u6587\u4ef6\uff08\u6743\u91cd\u548c\u914d\u7f6e\uff09\u3002<\/li>\n\n\n\n<li>\u521d\u6b21\u4f7f\u7528\u65f6\uff0c\u6a21\u578b\u6587\u4ef6\u7684\u8bbe\u7f6e\u6bd4 Haar \u65b9\u6cd5\u7a0d\u590d\u6742\u3002<\/li>\n<\/ul>\n\n\n\n<h4 class=\"wp-block-heading\">\u51c6\u5907\u5de5\u4f5c\uff1a\u4e0b\u8f7d\u6a21\u578b\u6587\u4ef6<\/h4>\n\n\n\n<p class=\"wp-block-paragraph\">\u4f60\u9700\u8981\u4e0b\u8f7d\u4e24\u4e2a\u6587\u4ef6\uff1a<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li><strong>\u6a21\u578b\u914d\u7f6e\u6587\u4ef6 (<code>.prototxt<\/code>)<\/strong>:&nbsp;<a href=\"https:\/\/raw.githubusercontent.com\/opencv\/opencv_3rdparty\/dnn_samples_face_detector_20170830\/deploy.prototxt\">deploy.prototxt.txt<\/a>&nbsp;(\u4e0b\u8f7d\u540e\u53bb\u6389&nbsp;<code>.txt<\/code>&nbsp;\u540e\u7f00)<\/li>\n\n\n\n<li><strong>\u6a21\u578b\u6743\u91cd\u6587\u4ef6 (<code>.caffemodel<\/code>)<\/strong>:&nbsp;<a href=\"https:\/\/github.com\/opencv\/opencv_3rdparty\/raw\/dnn_samples_face_detector_20170830\/res10_300x300_ssd_iter_140000.caffemodel\">res10_300x300_ssd_iter_140000.caffemodel<\/a><\/li>\n<\/ol>\n\n\n\n<p class=\"wp-block-paragraph\">\u5c06\u8fd9\u4e24\u4e2a\u6587\u4ef6\u548c\u4f60\u7684 Python \u811a\u672c\u653e\u5728\u540c\u4e00\u4e2a\u76ee\u5f55\u4e0b\u3002<\/p>\n\n\n\n<h4 class=\"wp-block-heading\">\u5b9e\u73b0\u6b65\u9aa4<\/h4>\n\n\n\n<ol class=\"wp-block-list\">\n<li><strong>\u52a0\u8f7d DNN \u6a21\u578b<\/strong>\uff1a\u4f7f\u7528&nbsp;<code>cv2.dnn.readNetFromCaffe()<\/code>&nbsp;\u52a0\u8f7d\u6a21\u578b\u3002<\/li>\n\n\n\n<li><strong>\u9884\u5904\u7406\u56fe\u50cf<\/strong>\uff1a\u5c06\u56fe\u50cf\u8f6c\u6362\u4e3a\u6a21\u578b\u671f\u671b\u7684\u8f93\u5165\u683c\u5f0f\uff08\u5c3a\u5bf8\u3001\u7f29\u653e\u3001\u901a\u9053\u987a\u5e8f\u7b49\uff09\u3002<\/li>\n\n\n\n<li><strong>\u524d\u5411\u4f20\u64ad<\/strong>\uff1a\u8c03\u7528&nbsp;<code>net.forward()<\/code>&nbsp;\u8fdb\u884c\u63a8\u7406\uff0c\u5f97\u5230\u68c0\u6d4b\u7ed3\u679c\u3002<\/li>\n\n\n\n<li><strong>\u89e3\u6790\u7ed3\u679c<\/strong>\uff1a\u5904\u7406\u7f51\u7edc\u8f93\u51fa\u7684\u7f6e\u4fe1\u5ea6\u548c\u8fb9\u754c\u6846\u3002<\/li>\n\n\n\n<li><strong>\u7ed8\u5236\u7ed3\u679c<\/strong>\uff1a\u7b5b\u9009\u51fa\u9ad8\u7f6e\u4fe1\u5ea6\u7684\u68c0\u6d4b\u6846\u5e76\u7ed8\u5236\u3002<\/li>\n<\/ol>\n\n\n\n<h4 class=\"wp-block-heading\">\u4ee3\u7801\u793a\u4f8b<\/h4>\n\n\n\n<pre class=\"wp-block-code\"><code>import cv2\nimport numpy as np\n\n<em># 1. \u52a0\u8f7d\u9884\u8bad\u7ec3\u7684 Caffe \u6a21\u578b<\/em>\nmodel_file = 'res10_300x300_ssd_iter_140000.caffemodel'\nconfig_file = 'deploy.prototxt'\n\nnet = cv2.dnn.readNetFromCaffe(config_file, model_file)\n\n<em># 2. \u8bfb\u53d6\u56fe\u50cf<\/em>\nimage_path = 'your_image.jpg'\nimage = cv2.imread(image_path)\n\nif image is None:\n    print(f\"\u9519\u8bef\uff1a\u65e0\u6cd5\u5728\u8def\u5f84 '{image_path}' \u627e\u5230\u56fe\u50cf\u3002\")\n    exit()\n\n<em># \u83b7\u53d6\u56fe\u50cf\u7684\u5c3a\u5bf8<\/em>\n(h, w) = image.shape&#91;:2]\n\n<em># 3. \u9884\u5904\u7406\u56fe\u50cf\uff0c\u6784\u5efa\u4e00\u4e2a \"blob\"<\/em>\n<em># cv2.dnn.blobFromImage \u53c2\u6570\u8bf4\u660e\uff1a<\/em>\n<em># image: \u8f93\u5165\u56fe\u50cf<\/em>\n<em># scalefactor: \u56fe\u50cf\u50cf\u7d20\u503c\u7684\u7f29\u653e\u56e0\u5b50<\/em>\n<em># size: \u6a21\u578b\u671f\u671b\u7684\u8f93\u5165\u5c3a\u5bf8<\/em>\n<em># mean: \u4ece\u6bcf\u4e2a\u901a\u9053\u4e2d\u51cf\u53bb\u7684\u5747\u503c<\/em>\n<em># swapRB: \u662f\u5426\u4ea4\u6362\u7ea2\u8272\u548c\u84dd\u8272\u901a\u9053 (OpenCV \u9ed8\u8ba4\u662f BGR, Caffe \u6a21\u578b\u901a\u5e38\u9700\u8981 RGB)<\/em>\nblob = cv2.dnn.blobFromImage(cv2.resize(image, (300, 300)), 1.0, (300, 300), (104.0, 177.0, 123.0), swapRB=False)\n\n<em># 4. \u5c06 blob \u8f93\u5165\u7f51\u7edc\u5e76\u8fdb\u884c\u524d\u5411\u4f20\u64ad<\/em>\nnet.setInput(blob)\ndetections = net.forward()\n\n<em># 5. \u89e3\u6790\u7ed3\u679c\u5e76\u7ed8\u5236<\/em>\n<em># detections \u7684\u7ef4\u5ea6\u662f (1, 1, N, 7)\uff0c\u5176\u4e2d N \u662f\u68c0\u6d4b\u5230\u7684\u4eba\u8138\u6570\u91cf<\/em>\n<em># \u6bcf\u4e2a\u68c0\u6d4b\u5411\u91cf\u7684\u683c\u5f0f\u662f &#91;batch_id, class_id, confidence, left, top, right, bottom]<\/em>\nfor i in range(detections.shape&#91;2]):\n    confidence = detections&#91;0, 0, i, 2]\n\n    <em># \u8fc7\u6ee4\u6389\u4f4e\u7f6e\u4fe1\u5ea6\u7684\u68c0\u6d4b<\/em>\n    if confidence &gt; 0.7:  <em># \u7f6e\u4fe1\u5ea6\u9608\u503c\u8bbe\u4e3a 0.7<\/em>\n        <em># \u8ba1\u7b97\u8fb9\u754c\u6846\u7684\u5750\u6807 (\u6ce8\u610f\uff1a\u9700\u8981\u5c06\u5750\u6807\u7f29\u653e\u56de\u539f\u59cb\u56fe\u50cf\u5c3a\u5bf8)<\/em>\n        box = detections&#91;0, 0, i, 3:7] * np.array(&#91;w, h, w, h])\n        (startX, startY, endX, endY) = box.astype(\"int\")\n\n        <em># \u7ed8\u5236\u8fb9\u754c\u6846\u548c\u7f6e\u4fe1\u5ea6<\/em>\n        text = f\"{confidence * 100:.2f}%\"\n        cv2.rectangle(image, (startX, startY), (endX, endY), (0, 255, 0), 2)\n        y = startY - 10 if startY - 10 &gt; 10 else startY + 10\n        cv2.putText(image, text, (startX, y), cv2.FONT_HERSHEY_SIMPLEX, 0.45, (0, 255, 0), 2)\n\n<em># \u663e\u793a\u7ed3\u679c\u56fe\u50cf<\/em>\ncv2.imshow('DNN Face Detection', image)\n\n<em># \u7b49\u5f85\u6309\u952e\uff0c\u7136\u540e\u5173\u95ed\u6240\u6709\u7a97\u53e3<\/em>\ncv2.waitKey(0)\ncv2.destroyAllWindows()\n<\/code><\/pre>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h3 class=\"wp-block-heading\">\u7efc\u5408\u5e94\u7528\uff1a\u4f7f\u7528 DNN \u8fdb\u884c\u4eba\u8138\u68c0\u6d4b + Dlib \u8fdb\u884c\u4eba\u8138\u5173\u952e\u70b9\u68c0\u6d4b<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">\u8fd9\u662f\u4e00\u4e2a\u975e\u5e38\u5f3a\u5927\u4e14\u5b9e\u7528\u7684\u7ec4\u5408\u3002\u6211\u4eec\u5229\u7528 OpenCV DNN \u6a21\u5757\u5feb\u901f\u51c6\u786e\u5730\u5b9a\u4f4d\u4eba\u8138\uff0c\u7136\u540e\u4f7f\u7528 Dlib \u5728\u8fd9\u4e9b\u7cbe\u786e\u7684\u4eba\u8138\u533a\u57df\u5185\u8fdb\u884c\u5173\u952e\u70b9\u68c0\u6d4b\u3002\u8fd9\u6837\u53ef\u4ee5\u7ed3\u5408\u4e24\u8005\u7684\u4f18\u70b9\uff1a<strong>\u5feb\u901f\u3001\u51c6\u786e\u7684\u68c0\u6d4b + \u7cbe\u7ec6\u7684\u5173\u952e\u70b9\u5b9a\u4f4d<\/strong>\u3002<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>import cv2\nimport dlib\nimport numpy as np\n\n<em># --- \u52a0\u8f7d\u6a21\u578b ---<\/em>\n<em># 1. OpenCV DNN \u4eba\u8138\u68c0\u6d4b\u6a21\u578b<\/em>\nmodel_file = 'res10_300x300_ssd_iter_140000.caffemodel'\nconfig_file = 'deploy.prototxt'\ndnn_net = cv2.dnn.readNetFromCaffe(config_file, model_file)\n\n<em># 2. Dlib \u5173\u952e\u70b9\u68c0\u6d4b\u6a21\u578b<\/em>\npredictor_path = 'shape_predictor_68_face_landmarks.dat'\ndlib_predictor = dlib.shape_predictor(predictor_path)\n\n\n<em># --- \u8bfb\u53d6\u5e76\u5904\u7406\u56fe\u50cf ---<\/em>\nimage_path = 'your_image.jpg'\nimage = cv2.imread(image_path)\nif image is None:\n    print(f\"\u9519\u8bef\uff1a\u65e0\u6cd5\u5728\u8def\u5f84 '{image_path}' \u627e\u5230\u56fe\u50cf\u3002\")\n    exit()\n\n(h, w) = image.shape&#91;:2]\ngray_image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)\n\n<em># --- \u6b65\u9aa4 1: \u4f7f\u7528 OpenCV DNN \u68c0\u6d4b\u4eba\u8138 ---<\/em>\nblob = cv2.dnn.blobFromImage(cv2.resize(image, (300, 300)), 1.0, (300, 300), (104.0, 177.0, 123.0), swapRB=False)\ndnn_net.setInput(blob)\ndetections = dnn_net.forward()\n\n<em># --- \u6b65\u9aa4 2: \u904d\u5386\u68c0\u6d4b\u7ed3\u679c\uff0c\u5e76\u4f7f\u7528 Dlib \u9884\u6d4b\u5173\u952e\u70b9 ---<\/em>\nfor i in range(detections.shape&#91;2]):\n    confidence = detections&#91;0, 0, i, 2]\n    if confidence &gt; 0.7:\n        <em># \u83b7\u53d6 DNN \u68c0\u6d4b\u6846\u7684\u5750\u6807<\/em>\n        box = detections&#91;0, 0, i, 3:7] * np.array(&#91;w, h, w, h])\n        (startX, startY, endX, endY) = box.astype(\"int\")\n\n        <em># --- \u5173\u952e\u8f6c\u6362\uff1a\u5c06 OpenCV \u7684\u77e9\u5f62\u6846 \u8f6c\u6362\u4e3a Dlib \u7684\u77e9\u5f62\u6846 ---<\/em>\n        <em># Dlib \u7684 rectangle \u683c\u5f0f\u662f (left, top, right, bottom)<\/em>\n        dlib_rect = dlib.rectangle(startX, startY, endX, endY)\n        \n        <em># \u4f7f\u7528 Dlib \u9884\u6d4b\u5668\u83b7\u53d6\u5173\u952e\u70b9<\/em>\n        landmarks = dlib_predictor(gray_image, dlib_rect)\n        \n        <em># --- \u7ed8\u5236\u7ed3\u679c ---<\/em>\n        <em># \u7ed8\u5236 DNN \u68c0\u6d4b\u6846<\/em>\n        cv2.rectangle(image, (startX, startY), (endX, endY), (0, 0, 255), 2) <em># \u7528\u7ea2\u8272\u8868\u793a DNN \u68c0\u6d4b\u6846<\/em>\n        \n        <em># \u7ed8\u5236 Dlib \u5173\u952e\u70b9<\/em>\n        for n in range(68):\n            x = landmarks.part(n).x\n            y = landmarks.part(n).y\n            cv2.circle(image, (x, y), 2, (0, 255, 0), -1) <em># \u7528\u7eff\u8272\u8868\u793a\u5173\u952e\u70b9<\/em>\n\n<em># --- \u663e\u793a\u6700\u7ec8\u7ed3\u679c ---<\/em>\ncv2.imshow('DNN + Dlib Face Detection and Landmarks', image)\ncv2.waitKey(0)\ncv2.destroyAllWindows()\n<\/code><\/pre>\n\n\n\n<h3 class=\"wp-block-heading\">\u603b\u7ed3\u4e0e\u9009\u62e9\u5efa\u8bae<\/h3>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><th>\u65b9\u6cd5<\/th><th>\u4f18\u70b9<\/th><th>\u7f3a\u70b9<\/th><th>\u9002\u7528\u573a\u666f<\/th><\/tr><\/thead><tbody><tr><td><strong>Haar \u7ea7\u8054<\/strong><\/td><td>\u901f\u5ea6\u5feb\uff0c\u65e0\u989d\u5916\u4f9d\u8d56\uff0c\u7b80\u5355\u6613\u7528<\/td><td>\u7cbe\u5ea6\u4f4e\uff0c\u6613\u53d7\u5e72\u6270\uff0c\u529f\u80fd\u5355\u4e00<\/td><td>\u5bf9\u7cbe\u5ea6\u8981\u6c42\u4e0d\u9ad8\u7684\u5feb\u901f\u539f\u578b\u3001\u5d4c\u5165\u5f0f\u8bbe\u5907\u3001\u5b66\u4e60\u5165\u95e8<\/td><\/tr><tr><td><strong>Dlib \u5173\u952e\u70b9<\/strong><\/td><td>\u7cbe\u5ea6\u6781\u9ad8\uff0c\u63d0\u4f9b68\u4e2a\u6807\u51c6\u70b9<\/td><td>\u901f\u5ea6\u8f83\u6162\uff0c\u9700\u5b89\u88c5Dlib<\/td><td>\u9700\u8981\u9ad8\u7cbe\u5ea6\u7279\u5f81\u70b9\u5206\u6790\u7684\u5e94\u7528\uff0c\u5982\u4eba\u8138\u5bf9\u9f50\u3001\u8868\u60c5\u8bc6\u522b\u3001AR\u7279\u6548<\/td><\/tr><tr><td><strong>OpenCV DNN<\/strong><\/td><td><strong>\u7cbe\u5ea6\u9ad8\uff0c\u901f\u5ea6\u5feb\uff0c\u73b0\u4ee3\u6807\u51c6<\/strong><\/td><td>\u9700\u4e0b\u8f7d\u6a21\u578b\u6587\u4ef6\uff0c\u914d\u7f6e\u7a0d\u590d\u6742<\/td><td><strong>\u7edd\u5927\u591a\u6570\u73b0\u4ee3\u5e94\u7528\u7684\u9996\u9009<\/strong>\uff0c\u5982\u5b9e\u65f6\u89c6\u9891\u6d41\u5206\u6790\u3001\u5b89\u9632\u76d1\u63a7\u3001\u4eba\u8138\u8bc6\u522b\u7cfb\u7edf<\/td><\/tr><tr><td><strong>DNN + Dlib<\/strong><\/td><td>\u7ed3\u5408\u4e86\u901f\u5ea6\u3001\u7cbe\u5ea6\u548c\u7cbe\u7ec6\u7279\u5f81\u70b9<\/td><td>\u4f9d\u8d56\u4e24\u4e2a\u5e93\uff0c\u6d41\u7a0b\u7a0d\u590d\u6742<\/td><td>\u9700\u8981\u540c\u65f6\u8fdb\u884c\u9ad8\u7cbe\u5ea6\u68c0\u6d4b\u548c\u7cbe\u7ec6\u7279\u5f81\u5206\u6790\u7684\u9ad8\u7aef\u5e94\u7528\uff0c\u5982\u9ad8\u7ea7\u7f8e\u989c\u30013D\u4eba\u8138\u91cd\u5efa<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>\u7ed9\u4f60\u7684\u5efa\u8bae<\/strong>\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>\u521d\u5b66\u8005<\/strong>\uff1a\u4ece&nbsp;<strong>Haar \u7ea7\u8054<\/strong>&nbsp;\u5f00\u59cb\uff0c\u7406\u89e3\u4eba\u8138\u68c0\u6d4b\u7684\u57fa\u672c\u6d41\u7a0b\u3002<\/li>\n\n\n\n<li><strong>\u5b9e\u9645\u9879\u76ee<\/strong>\uff1a\u76f4\u63a5\u4f7f\u7528&nbsp;<strong>OpenCV DNN<\/strong>&nbsp;\u6a21\u5757\u8fdb\u884c\u4eba\u8138\u68c0\u6d4b\uff0c\u5b83\u5728\u6027\u80fd\u548c\u6613\u7528\u6027\u4e0a\u662f\u6700\u4f73\u9009\u62e9\u3002<\/li>\n\n\n\n<li><strong>\u9700\u8981\u5173\u952e\u70b9<\/strong>\uff1a\u5728&nbsp;<strong>OpenCV DNN<\/strong>&nbsp;\u68c0\u6d4b\u5230\u7684\u4eba\u8138\u57fa\u7840\u4e0a\uff0c\u518d\u4f7f\u7528&nbsp;<strong>Dlib<\/strong>&nbsp;\u8fdb\u884c\u5173\u952e\u70b9\u68c0\u6d4b\uff0c\u8fd9\u662f\u76ee\u524d\u4e1a\u754c\u975e\u5e38\u6210\u719f\u548c\u9ad8\u6548\u7684\u65b9\u6848\u3002<\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>\u8fd9\u91cc\u6211\u4eec\u4e3b\u8981\u5173\u6ce8\u4e24\u79cd\u6700\u5e38\u89c1\u7684\u201c\u7279\u5f81\u8bc6\u522b\u201d\uff1a<\/p>\n<p>\u4eba\u8138\u68c0\u6d4b\uff1a\u5728\u56fe\u50cf\u6216\u89c6\u9891\u4e2d\u627e\u5230\u4eba\u8138\u7684\u4f4d\u7f6e\u548c\u5927\u5c0f\uff0c\u5e76\u7528\u4e00\u4e2a\u65b9\u6846\u6846\u51fa\u6765\u3002\u8fd9\u662f\u6240\u6709\u540e\u7eed\u64cd\u4f5c\u7684\u57fa\u7840\u3002<\/p>\n<p>\u4eba\u8138\u5173\u952e\u70b9\u68c0\u6d4b\uff1a\u5728\u68c0\u6d4b\u5230\u7684\u4eba\u8138\u4e0a\uff0c\u5b9a\u4f4d\u51fa\u5173\u952e\u7684\u7279\u5f81\u70b9\uff0c\u5982\u773c\u775b\u3001\u9f3b\u5b50\u3001\u5634\u5df4\u3001\u7709\u6bdb\u7684\u8f6e\u5ed3\u7b49\u3002\u8fd9\u901a\u5e38\u88ab\u79f0\u4e3a\u201c\u4eba\u8138\u5bf9\u9f50\u201d\u6216\u201c\u7279\u5f81\u70b9\u6807\u8bb0\u201d\u3002<\/p>\n<p>\u6211\u4eec\u5c06\u4ece\u6700\u57fa\u7840\u3001\u6700\u7ecf\u5178\u7684\u65b9\u6cd5\u5f00\u59cb\uff0c\u9010\u6b65\u8fc7\u6e21\u5230\u66f4\u73b0\u4ee3\u3001\u66f4\u7cbe\u786e\u7684<\/p>\n","protected":false},"author":1,"featured_media":207,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_kadence_starter_templates_imported_post":false,"footnotes":""},"categories":[4,6,5],"tags":[],"class_list":["post-35","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-4","category-6","category-5"],"_links":{"self":[{"href":"https:\/\/dreamrunrun.com\/blog\/index.php?rest_route=\/wp\/v2\/posts\/35","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/dreamrunrun.com\/blog\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/dreamrunrun.com\/blog\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/dreamrunrun.com\/blog\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/dreamrunrun.com\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=35"}],"version-history":[{"count":2,"href":"https:\/\/dreamrunrun.com\/blog\/index.php?rest_route=\/wp\/v2\/posts\/35\/revisions"}],"predecessor-version":[{"id":137,"href":"https:\/\/dreamrunrun.com\/blog\/index.php?rest_route=\/wp\/v2\/posts\/35\/revisions\/137"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/dreamrunrun.com\/blog\/index.php?rest_route=\/wp\/v2\/media\/207"}],"wp:attachment":[{"href":"https:\/\/dreamrunrun.com\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=35"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/dreamrunrun.com\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=35"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/dreamrunrun.com\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=35"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}