{"id":945961,"date":"2024-12-26T23:30:58","date_gmt":"2024-12-26T15:30:58","guid":{"rendered":"https:\/\/docs.pingcode.com\/ask\/ask-ask\/945961.html"},"modified":"2024-12-26T23:31:00","modified_gmt":"2024-12-26T15:31:00","slug":"python%e5%a6%82%e4%bd%95%e6%98%be%e7%a4%baminist","status":"publish","type":"post","link":"https:\/\/docs.pingcode.com\/ask\/945961.html","title":{"rendered":"python\u5982\u4f55\u663e\u793aminist"},"content":{"rendered":"

\"python\u5982\u4f55\u663e\u793aminist\"<\/p>\n

\u8981\u5728Python\u4e2d\u663e\u793aMNIST\u6570\u636e\u96c6\uff0c\u53ef\u4ee5\u4f7f\u7528\u51e0\u4e2a\u5e38\u7528\u7684\u5e93\uff0c\u5982TensorFlow\u3001Keras\u548cMatplotlib\u7b49\u3002MNIST\u662f\u4e00\u4e2a\u5927\u578b\u7684\u624b\u5199\u6570\u5b57\u6570\u636e\u5e93\uff0c\u5e38\u7528\u4e8e\u8bad\u7ec3\u5404\u79cd\u56fe\u50cf\u5904\u7406\u7cfb\u7edf\u3002\u8981\u663e\u793aMNIST\u6570\u636e\u96c6\uff0c\u4f60\u53ef\u4ee5\u4f7f\u7528TensorFlow\u6216Keras\u52a0\u8f7d\u6570\u636e\uff0c\u5e76\u4f7f\u7528Matplotlib\u8fdb\u884c\u53ef\u89c6\u5316\u3002<\/strong>\u5176\u4e2d\uff0c\u4f7f\u7528Keras\u52a0\u8f7d\u6570\u636e\u662f\u76f8\u5bf9\u7b80\u5355\u4e14\u76f4\u89c2\u7684\u65b9\u6cd5\u3002\u63a5\u4e0b\u6765\uff0c\u6211\u5c06\u8be6\u7ec6\u63cf\u8ff0\u5982\u4f55\u5728Python\u4e2d\u663e\u793aMNIST\u6570\u636e\u96c6\u3002<\/p>\n<\/p>\n

\u4e00\u3001\u52a0\u8f7dMNIST\u6570\u636e\u96c6<\/p>\n<\/p>\n

MNIST\u6570\u636e\u96c6\u53ef\u4ee5\u901a\u8fc7Keras\u5e93\u8f7b\u677e\u52a0\u8f7d\u3002Keras\u662f\u4e00\u4e2a\u9ad8\u7ea7\u795e\u7ecf\u7f51\u7edcAPI\uff0c\u80fd\u591f\u5feb\u901f\u52a0\u8f7d\u548c\u9884\u5904\u7406\u5e38\u7528\u7684\u6570\u636e\u96c6\u3002Keras\u5305\u542b\u5728TensorFlow\u4e2d\uff0c\u56e0\u6b64\u53ea\u9700\u8981\u5b89\u88c5TensorFlow\u5373\u53ef\u4f7f\u7528Keras\u3002<\/p>\n<\/p>\n

from tensorflow.keras.datasets import mnist<\/p>\n
\u52a0\u8f7dMNIST\u6570\u636e\u96c6<\/strong><\/h2>\n
(trAI<\/a>n_images, train_labels), (test_images, test_labels) = mnist.load_data()<\/p>\n
<\/code><\/pre>\n<\/p>\n
\u5728\u8fd9\u6bb5\u4ee3\u7801\u4e2d\uff0cmnist.load_data()<\/code>\u51fd\u6570\u5c06MNIST\u6570\u636e\u96c6\u5206\u4e3a\u8bad\u7ec3\u96c6\u548c\u6d4b\u8bd5\u96c6\u3002\u8bad\u7ec3\u96c6\u5305\u542b60,000\u4e2a\u6837\u672c\uff0c\u6d4b\u8bd5\u96c6\u5305\u542b10,000\u4e2a\u6837\u672c\u3002\u6bcf\u4e2a\u6837\u672c\u662f\u4e00\u4e2a28×28\u7684\u7070\u5ea6\u56fe\u50cf\uff0c\u4ee3\u8868\u624b\u5199\u6570\u5b57\u3002<\/p>\n<\/p>\n
\u4e8c\u3001\u4e86\u89e3\u6570\u636e\u96c6\u7ed3\u6784<\/p>\n<\/p>\n
\u5728\u663e\u793aMNIST\u6570\u636e\u96c6\u4e4b\u524d\uff0c\u4e86\u89e3\u6570\u636e\u96c6\u7684\u7ed3\u6784\u662f\u5f88\u91cd\u8981\u7684\u3002MNIST\u6570\u636e\u96c6\u4e2d\u7684\u6bcf\u4e2a\u6837\u672c\u90fd\u662f\u4e00\u4e2a28×28\u7684\u4e8c\u7ef4\u6570\u7ec4\uff0c\u8868\u793a\u4e00\u4e2a\u7070\u5ea6\u503c\u56fe\u50cf\u3002\u6807\u7b7e\u662f\u4e00\u4e2a\u6574\u6570\uff0c\u8868\u793a\u56fe\u50cf\u4e2d\u624b\u5199\u6570\u5b57\u7684\u771f\u5b9e\u503c\u3002<\/p>\n<\/p>\n
# \u67e5\u770b\u6570\u636e\u96c6\u7684\u5f62\u72b6<\/p>\n
print('\u8bad\u7ec3\u6570\u636e\u96c6\u7684\u5f62\u72b6:', train_images.shape)<\/p>\n
print('\u8bad\u7ec3\u6807\u7b7e\u7684\u5f62\u72b6:', train_labels.shape)<\/p>\n
print('\u6d4b\u8bd5\u6570\u636e\u96c6\u7684\u5f62\u72b6:', test_images.shape)<\/p>\n
print('\u6d4b\u8bd5\u6807\u7b7e\u7684\u5f62\u72b6:', test_labels.shape)<\/p>\n
\u8f93\u51fa\u67d0\u4e2a\u6837\u672c\u7684\u50cf\u7d20\u503c\u548c\u6807\u7b7e<\/strong><\/h2>\n
sample_index = 0<\/p>\n
print('\u6837\u672c\u7684\u50cf\u7d20\u503c:\\n', train_images[sample_index])<\/p>\n
print('\u6837\u672c\u7684\u6807\u7b7e:', train_labels[sample_index])<\/p>\n
<\/code><\/pre>\n<\/p>\n
\u8fd9\u4e9b\u4ee3\u7801\u5c06\u8f93\u51fa\u8bad\u7ec3\u548c\u6d4b\u8bd5\u6570\u636e\u96c6\u7684\u5f62\u72b6\uff0c\u4ee5\u53ca\u4e00\u4e2a\u6837\u672c\u7684\u50cf\u7d20\u503c\u548c\u6807\u7b7e\u3002\u901a\u8fc7\u8fd9\u4e9b\u4fe1\u606f\uff0c\u6211\u4eec\u53ef\u4ee5\u786e\u8ba4\u6570\u636e\u96c6\u7684\u7ed3\u6784\u548c\u5185\u5bb9\u3002<\/p>\n<\/p>\n
\u4e09\u3001\u663e\u793aMNIST\u56fe\u50cf<\/p>\n<\/p>\n
\u4f7f\u7528Matplotlib\u5e93\u53ef\u4ee5\u8f7b\u677e\u5730\u5c06MNIST\u6570\u636e\u96c6\u4e2d\u7684\u56fe\u50cf\u663e\u793a\u51fa\u6765\u3002Matplotlib\u662fPython\u4e2d\u4e00\u4e2a\u5f3a\u5927\u7684\u7ed8\u56fe\u5e93\uff0c\u652f\u6301\u591a\u79cd\u56fe\u5f62\u548c\u56fe\u50cf\u7684\u53ef\u89c6\u5316\u3002<\/p>\n<\/p>\n
import matplotlib.pyplot as plt<\/p>\n
\u663e\u793a\u67d0\u4e2aMNIST\u56fe\u50cf<\/strong><\/h2>\n
def display_sample(index):<\/p>\n
    plt.imshow(train_images[index], cmap='gray')<\/p>\n
    plt.title(f'\u6807\u7b7e: {train_labels[index]}')<\/p>\n
    plt.show()<\/p>\n
\u663e\u793a\u7b2c\u4e00\u4e2a\u6837\u672c<\/strong><\/h2>\n
display_sample(0)<\/p>\n
<\/code><\/pre>\n<\/p>\n
\u5728\u8fd9\u6bb5\u4ee3\u7801\u4e2d\uff0cdisplay_sample<\/code>\u51fd\u6570\u7528\u4e8e\u663e\u793a\u6307\u5b9a\u7d22\u5f15\u7684MNIST\u56fe\u50cf\u3002plt.imshow<\/code>\u51fd\u6570\u7528\u4e8e\u663e\u793a\u56fe\u50cf\uff0ccmap='gray'<\/code>\u8868\u793a\u4f7f\u7528\u7070\u5ea6\u989c\u8272\u6620\u5c04\u3002plt.title<\/code>\u7528\u4e8e\u5728\u56fe\u50cf\u4e0a\u65b9\u663e\u793a\u6807\u7b7e\u3002<\/p>\n<\/p>\n
\u56db\u3001\u6279\u91cf\u663e\u793aMNIST\u56fe\u50cf<\/p>\n<\/p>\n
\u4e3a\u4e86\u66f4\u597d\u5730\u4e86\u89e3\u6570\u636e\u96c6\u7684\u591a\u6837\u6027\uff0c\u53ef\u4ee5\u4e00\u6b21\u663e\u793a\u591a\u4e2aMNIST\u56fe\u50cf\u3002<\/p>\n<\/p>\n
# \u6279\u91cf\u663e\u793aMNIST\u56fe\u50cf<\/p>\n
def display_samples(indices):<\/p>\n
    plt.figure(figsize=(10, 10))<\/p>\n
    for i, index in enumerate(indices):<\/p>\n
        plt.subplot(1, len(indices), i + 1)<\/p>\n
        plt.imshow(train_images[index], cmap='gray')<\/p>\n
        plt.title(f'\u6807\u7b7e: {train_labels[index]}')<\/p>\n
        plt.axis('off')<\/p>\n
    plt.show()<\/p>\n
\u663e\u793a\u524d5\u4e2a\u6837\u672c<\/strong><\/h2>\n
display_samples([0, 1, 2, 3, 4])<\/p>\n
<\/code><\/pre>\n<\/p>\n
\u5728\u8fd9\u6bb5\u4ee3\u7801\u4e2d\uff0cdisplay_samples<\/code>\u51fd\u6570\u7528\u4e8e\u663e\u793a\u591a\u4e2aMNIST\u56fe\u50cf\u3002\u6211\u4eec\u4f7f\u7528plt.subplot<\/code>\u51fd\u6570\u5728\u4e00\u884c\u4e2d\u7ed8\u5236\u591a\u5e45\u56fe\u50cf\uff0c\u5e76\u4f7f\u7528plt.axis('off')<\/code>\u9690\u85cf\u5750\u6807\u8f74\u3002<\/p>\n<\/p>\n
\u4e94\u3001\u6570\u636e\u9884\u5904\u7406<\/p>\n<\/p>\n
\u5728\u5c06MNIST\u6570\u636e\u96c6\u7528\u4e8e\u8bad\u7ec3\u795e\u7ecf\u7f51\u7edc\u4e4b\u524d\uff0c\u901a\u5e38\u9700\u8981\u8fdb\u884c\u6570\u636e\u9884\u5904\u7406\u3002\u5e38\u89c1\u7684\u9884\u5904\u7406\u6b65\u9aa4\u5305\u62ec\u5f52\u4e00\u5316\u548c\u5f62\u72b6\u53d8\u6362\u3002<\/p>\n<\/p>\n
# \u5f52\u4e00\u5316\u50cf\u7d20\u503c<\/p>\n
train_images = train_images \/ 255.0<\/p>\n
test_images = test_images \/ 255.0<\/p>\n
\u5c06\u56fe\u50cf\u5f62\u72b6\u53d8\u4e3a (\u6837\u672c\u6570, \u9ad8, \u5bbd, \u901a\u9053\u6570) \u7684\u683c\u5f0f<\/strong><\/h2>\n
train_images = train_images.reshape((train_images.shape[0], 28, 28, 1))<\/p>\n
test_images = test_images.reshape((test_images.shape[0], 28, 28, 1))<\/p>\n
<\/code><\/pre>\n<\/p>\n
\u901a\u8fc7\u5c06\u50cf\u7d20\u503c\u9664\u4ee5255.0\uff0c\u6211\u4eec\u5c06\u5176\u5f52\u4e00\u5316\u5230[0, 1]\u7684\u8303\u56f4\u3002\u8fd9\u6837\u6709\u52a9\u4e8e\u63d0\u9ad8\u795e\u7ecf\u7f51\u7edc\u7684\u8bad\u7ec3\u6548\u679c\u3002\u5c06\u56fe\u50cf\u91cd\u5851\u4e3a\u56db\u7ef4\u6570\u7ec4\u662f\u4e3a\u4e86\u9002\u5e94\u5377\u79ef\u795e\u7ecf\u7f51\u7edc(CNN)\u7684\u8f93\u5165\u683c\u5f0f\u3002<\/p>\n<\/p>\n
\u516d\u3001\u6784\u5efa\u548c\u8bad\u7ec3\u6a21\u578b<\/p>\n<\/p>\n
\u5728\u5b8c\u6210\u6570\u636e\u9884\u5904\u7406\u540e\uff0c\u53ef\u4ee5\u4f7f\u7528Keras\u6784\u5efa\u548c\u8bad\u7ec3\u795e\u7ecf\u7f51\u7edc\u6a21\u578b\u3002\u4ee5\u4e0b\u662f\u4e00\u4e2a\u7b80\u5355\u7684\u5377\u79ef\u795e\u7ecf\u7f51\u7edc(CNN)\u6a21\u578b\uff1a<\/p>\n<\/p>\n
from tensorflow.keras.models import Sequential<\/p>\n
from tensorflow.keras.layers import Conv2D, MaxPooling2D, Flatten, Dense<\/p>\n
\u6784\u5efaCNN\u6a21\u578b<\/strong><\/h2>\n
model = Sequential([<\/p>\n
    Conv2D(32, (3, 3), activation='relu', input_shape=(28, 28, 1)),<\/p>\n
    MaxPooling2D((2, 2)),<\/p>\n
    Flatten(),<\/p>\n
    Dense(64, activation='relu'),<\/p>\n
    Dense(10, activation='softmax')<\/p>\n
])<\/p>\n
\u7f16\u8bd1\u6a21\u578b<\/strong><\/h2>\n
model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy'])<\/p>\n
\u8bad\u7ec3\u6a21\u578b<\/strong><\/h2>\n
model.fit(train_images, train_labels, epochs=5)<\/p>\n
<\/code><\/pre>\n<\/p>\n
\u8fd9\u4e2a\u6a21\u578b\u5305\u542b\u4e00\u4e2a\u5377\u79ef\u5c42\u3001\u4e00\u4e2a\u6700\u5927\u6c60\u5316\u5c42\u548c\u4e24\u4e2a\u5168\u8fde\u63a5\u5c42\u3002\u6211\u4eec\u4f7f\u7528adam<\/code>\u4f18\u5316\u5668\u548csparse_categorical_crossentropy<\/code>\u635f\u5931\u51fd\u6570\u8fdb\u884c\u7f16\u8bd1\uff0c\u5e76\u8bad\u7ec3\u6a21\u578b5\u4e2a\u5468\u671f\u3002<\/p>\n<\/p>\n
\u4e03\u3001\u8bc4\u4f30\u6a21\u578b<\/p>\n<\/p>\n
\u8bad\u7ec3\u5b8c\u6210\u540e\uff0c\u53ef\u4ee5\u5728\u6d4b\u8bd5\u96c6\u4e0a\u8bc4\u4f30\u6a21\u578b\u7684\u6027\u80fd\uff1a<\/p>\n<\/p>\n
# \u8bc4\u4f30\u6a21\u578b<\/p>\n
test_loss, test_acc = model.evaluate(test_images, test_labels)<\/p>\n
print(f'\u6d4b\u8bd5\u51c6\u786e\u7387: {test_acc}')<\/p>\n
<\/code><\/pre>\n<\/p>\n
\u8bc4\u4f30\u7ed3\u679c\u5c06\u663e\u793a\u6a21\u578b\u5728\u6d4b\u8bd5\u96c6\u4e0a\u7684\u51c6\u786e\u7387\u3002<\/p>\n<\/p>\n
\u901a\u8fc7\u4ee5\u4e0a\u6b65\u9aa4\uff0c\u4f60\u53ef\u4ee5\u6210\u529f\u52a0\u8f7d\u3001\u663e\u793aMNIST\u6570\u636e\u96c6\uff0c\u5e76\u5728\u5176\u4e0a\u6784\u5efa\u548c\u8bad\u7ec3\u4e00\u4e2a\u7b80\u5355\u7684\u795e\u7ecf\u7f51\u7edc\u6a21\u578b\u3002MNIST\u662f\u4e00\u4e2a\u5165\u95e8\u7ea7\u7684\u6570\u636e\u96c6\uff0c\u975e\u5e38\u9002\u5408\u7528\u4e8e\u5b66\u4e60\u548c\u5b9e\u9a8c\u3002\u5e0c\u671b\u8fd9\u7bc7\u6587\u7ae0\u5bf9\u4f60\u6709\u6240\u5e2e\u52a9\uff01<\/p>\n<\/p>\n
\u76f8\u5173\u95ee\u7b54FAQs\uff1a<\/strong><\/h2>\n
 \u5982\u4f55\u5728Python\u4e2d\u52a0\u8f7d\u548c\u663e\u793aMNIST\u6570\u636e\u96c6\uff1f<\/strong>
\u8981\u5728Python\u4e2d\u52a0\u8f7dMNIST\u6570\u636e\u96c6\uff0c\u60a8\u53ef\u4ee5\u4f7f\u7528tensorflow<\/code>\u6216keras<\/code>\u5e93\u3002\u4ee5\u4e0b\u662f\u4e00\u4e2a\u7b80\u5355\u7684\u793a\u4f8b\u4ee3\u7801\uff1a  <\/p>\n
import matplotlib.pyplot as plt\nfrom tensorflow.keras.datasets import mnist\n\n# \u52a0\u8f7dMNIST\u6570\u636e\u96c6\n(x_train, y_train), (x_test, y_test) = mnist.load_data()\n\n# \u663e\u793a\u7b2c\u4e00\u5f20\u56fe\u7247\nplt.imshow(x_train[0], cmap='gray')\nplt.title(f'Label: {y_train[0]}')\nplt.show()\n<\/code><\/pre>\n
\u6b64\u4ee3\u7801\u5c06\u52a0\u8f7dMNIST\u6570\u636e\u96c6\u5e76\u663e\u793a\u7b2c\u4e00\u5f20\u56fe\u50cf\u53ca\u5176\u6807\u7b7e\u3002<\/p>\n
\u4f7f\u7528\u54ea\u4e9b\u5e93\u53ef\u4ee5\u66f4\u65b9\u4fbf\u5730\u5904\u7406MNIST\u6570\u636e\u96c6\uff1f<\/strong>
tensorflow<\/code>\u548ckeras<\/code>\u662f\u5904\u7406MNIST\u6570\u636e\u96c6\u7684\u6d41\u884c\u9009\u62e9\u3002numpy<\/code>\u4e5f\u5e38\u7528\u4e8e\u6570\u636e\u5904\u7406\u548c\u8f6c\u6362\u3002matplotlib<\/code>\u5e93\u53ef\u4ee5\u5e2e\u52a9\u60a8\u53ef\u89c6\u5316\u6570\u636e\u3002\u60a8\u53ef\u4ee5\u901a\u8fc7pip install<\/code>\u547d\u4ee4\u8f7b\u677e\u5b89\u88c5\u8fd9\u4e9b\u5e93\uff0c\u4f8b\u5982\uff1a  <\/p>\n
pip install tensorflow keras matplotlib\n<\/code><\/pre>\n
\u5982\u4f55\u5bf9MNIST\u6570\u636e\u96c6\u8fdb\u884c\u9884\u5904\u7406\u4ee5\u63d0\u9ad8\u6a21\u578b\u7684\u6027\u80fd\uff1f<\/strong>
\u5bf9MNIST\u6570\u636e\u96c6\u8fdb\u884c\u9884\u5904\u7406\u53ef\u4ee5\u663e\u8457\u63d0\u9ad8\u6a21\u578b\u7684\u6027\u80fd\u3002\u5e38\u89c1\u7684\u9884\u5904\u7406\u6b65\u9aa4\u5305\u62ec\uff1a  <\/p>\n
    \n
  1. \u5f52\u4e00\u5316<\/strong>\uff1a\u5c06\u56fe\u50cf\u6570\u636e\u7f29\u653e\u52300\u52301\u4e4b\u95f4\uff0c\u901a\u5e38\u901a\u8fc7\u5c06\u50cf\u7d20\u503c\u9664\u4ee5255\u5b9e\u73b0\u3002  <\/li>\n
  2. \u91cd\u5851<\/strong>\uff1a\u5c06\u6bcf\u5f2028×28\u7684\u56fe\u50cf\u91cd\u5851\u4e3a\u4e00\u7ef4\u6570\u7ec4\uff0c\u4ee5\u4fbf\u8f93\u5165\u5230\u6a21\u578b\u4e2d\u3002  <\/li>\n
  3. \u72ec\u70ed\u7f16\u7801<\/strong>\uff1a\u5c06\u6807\u7b7e\u8f6c\u6362\u4e3a\u72ec\u70ed\u7f16\u7801\u683c\u5f0f\uff0c\u4ee5\u4fbf\u4e8e\u5206\u7c7b\u6a21\u578b\u7684\u8bad\u7ec3\u3002
    \u4ee3\u7801\u793a\u4f8b\uff1a<\/li>\n<\/ol>\n
    x_train = x_train.astype('float32') \/ 255\nx_test = x_test.astype('float32') \/ 255\ny_train = keras.utils.to_categorical(y_train, 10)\ny_test = keras.utils.to_categorical(y_test, 10)\n<\/code><\/pre>\n
    \u8fd9\u4e9b\u6b65\u9aa4\u5c06\u5e2e\u52a9\u60a8\u7684\u6a21\u578b\u66f4\u5feb\u5730\u6536\u655b\u5e76\u63d0\u9ad8\u51c6\u786e\u6027\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"\u8981\u5728Python\u4e2d\u663e\u793aMNIST\u6570\u636e\u96c6\uff0c\u53ef\u4ee5\u4f7f\u7528\u51e0\u4e2a\u5e38\u7528\u7684\u5e93\uff0c\u5982TensorFlow\u3001Keras\u548cMatplo […]","protected":false},"author":3,"featured_media":945967,"comment_status":"closed","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[37],"tags":[],"acf":[],"_links":{"self":[{"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/posts\/945961"}],"collection":[{"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/users\/3"}],"replies":[{"embeddable":true,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/comments?post=945961"}],"version-history":[{"count":"1","href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/posts\/945961\/revisions"}],"predecessor-version":[{"id":945969,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/posts\/945961\/revisions\/945969"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/media\/945967"}],"wp:attachment":[{"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/media?parent=945961"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/categories?post=945961"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/tags?post=945961"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}