js\u80fd\u5f00\u53d1exe\u7a0b\u5e8f\u5417<\/a> import Adam<\/p>\nnum_classes = 10<\/p>\n
input_shape = (28, 28, 1)<\/p>\n
batch_size = 128<\/p>\n
epochs = 5<\/p>\n
(x_train, y_train), (x_test, y_test) = keras.datasets.mnist.load_data()<\/p>\n
x_train = x_train.reshape(x_train.shape[0], 28, 28, 1).astype(‘float32’)<\/p>\n
x_test = x_test.reshape(x_test.shape[0], 28, 28, 1).astype(‘float32’)<\/p>\n
x_train \/= 255<\/p>\n
x_test \/= 255<\/p>\n
y_train = keras.utils.to_categorical(y_train, num_classes)<\/p>\n
y_test = keras.utils.to_categorical(y_test, num_classes)<\/p>\n
model = Sequential()<\/p>\n
model.add(Conv2D(32, kernel_size=(3, 3), activation=’relu’, input_shape=input_shape))<\/p>\n
model.add(MaxPooling2D(pool_size=(2, 2)))<\/p>\n
model.add(Conv2D(64, kernel_size=(3, 3), activation=’relu’))<\/p>\n
model.add(MaxPooling2D(pool_size=(2, 2)))<\/p>\n
model.add(Flatten())<\/p>\n
model.add(Dense(128, activation=’relu’))<\/p>\n
model.add(Dropout(0.5))<\/p>\n
model.add(Dense(num_classes, activation=’softmax’))<\/p>\n
model.compile(loss=keras.losses.categorical_crossentropy,<\/p>\n
optimizer=Adam(),<\/p>\n
metrics=[‘accuracy’])<\/p>\n
model.fit(x_train, y_train,<\/p>\n
batch_size=batch_size,<\/p>\n
epochs=epochs,<\/p>\n
verbose=1,<\/p>\n
validation_data=(x_test, y_test))<\/p>\n
model.save(‘keras_model.h5’)<\/p>\n
“`<\/p>\n
\u8fd0\u884c\u8fd9\u6bb5\u4ee3\u7801\u5c06\u521b\u5efa\u4e00\u4e2a\u8bad\u7ec3\u597d\u7684\u5377\u79ef\u795e\u7ecf\u7f51\u7edc\u6a21\u578b\uff0c\u5e76\u5c06\u5176\u4fdd\u5b58\u4e3a’H5’\u6587\u4ef6\u3002<\/p>\n
\u6b65\u9aa43\uff1a\u7f16\u5199Python\u811a\u672c\u4ee5\u52a0\u8f7d\u548c\u6267\u884c\u6a21\u578b<\/p>\n
\u5728\u8fd9\u4e00\u6b65\u4e2d\uff0c\u60a8\u9700\u8981\u7f16\u5199\u4e00\u4e2aPython\u811a\u672c\uff0c\u7528\u4e8e\u52a0\u8f7d\u4e4b\u524d\u521b\u5efa\u7684Keras\u6a21\u578b\uff0c\u5e76\u6267\u884c\u4e00\u4e9b\u64cd\u4f5c\u3002\u4f8b\u5982\uff0c\u9884\u6d4b\u624b\u5199\u6570\u5b57\u3002\u5c06\u4ee5\u4e0b\u4ee3\u7801\u4fdd\u5b58\u4e3a\u4e00\u4e2a\u6587\u4ef6\uff0c\u5982`keras_predict.py`\u3002<\/p>\n
“`python<\/p>\n
import numpy as np<\/p>\n
import keras <\/p>\n
from keras.models import load_model<\/p>\n
from keras.preprocessing import image<\/p>\n
# \u52a0\u8f7d\u6a21\u578b<\/p>\n
model = load_model(‘keras_model.h5’)<\/p>\n
# \u8bfb\u5165\u8981\u8bc6\u522b\u7684\u56fe\u7247<\/p>\n
img_path = ‘input_image.png’<\/p>\n
img = image.load_img(img_path, grayscale=True, target_size=(28, 28))<\/p>\n
# \u5c06\u8f93\u5165\u56fe\u7247\u8f6c\u6362\u4e3aNUMPy Array\uff0c\u5e76\u5f52\u4e00\u5316<\/p>\n
input_image = image.img_to_array(img)<\/p>\n
input_image = np.expand_dims(input_image, axis=0)<\/p>\n
input_image \/= 255<\/p>\n
# \u9884\u6d4b\u8f93\u5165\u56fe\u7247\u7684\u7c7b\u522b<\/p>\n
prediction = model.predict_classes(input_image)<\/p>\n
print(‘Prediction:’, prediction[0])<\/p>\n
“`<\/p>\n
\u8fd9\u4e2a\u811a\u672c\u5c06\u52a0\u8f7d\u4e00\u4e2a\u540d\u4e3a’input_image.png’\u7684\u56fe\u7247\uff0c\u901a\u8fc7\u8bad\u7ec3\u597d\u7684Keras\u6a21\u578b\u8fdb\u884c\u9884\u6d4b\uff0c\u5e76\u6253\u5370\u9884\u6d4b\u7ed3\u679c\u3002\u60a8\u53ef\u4ee5\u4f7f\u7528\u4efb\u4f55\u9002\u5f53\u7684\u624b\u5199\u6570\u5b57\u56fe\u7247\u6d4b\u8bd5\u6b64\u811a\u672c\uff0c\u53ea\u9700\u5c06\u56fe\u7247\u8def\u5f84\u66f4\u6539\u4e3a\u60a8\u81ea\u5df1\u7684\u56fe\u7247\u8def\u5f84\u3002<\/p>\n
\u6b65\u9aa44\uff1a\u5c06Python\u811a\u672c\u6253\u5305\u6210exe\u6587\u4ef6<\/p>\n
\u6700\u540e,\u6211\u4eec\u5c06\u4f7f\u7528PyInstaller\u5c06\u5176\u6253\u5305\u4e3a\u4e00\u4e2a\u72ec\u7acb\u7684exe\u6587\u4ef6\u3002\u8fd0\u884c\u4ee5\u4e0b\u547d\u4ee4\uff1a<\/p>\n
“\u8f6f\u4ef6\u6253\u5305\u8f6f\u4ef6\u63a8\u8350<\/a>`<\/p>\npyinstaller –onefile –add-data “keras_model.h5;.” keras_predict.py<\/p>\n
“`<\/p>\n
\u5b8c\u6210\u540e\uff0c\u60a8\u4f1a\u5728\u521b\u5efa\u7684’dist’\u6587\u4ef6\u5939\u4e0b\u627e\u5230\u4e00\u4e2a\u540d\u4e3a`keras_predict.exe`\u7684\u53ef\u6267\u884c\u6587\u4ef6\u3002\u5c06\u8fd9\u4e2a\u53ef\u6267\u884c\u6587\u4ef6\u4e0e\u4e4b\u524d\u4fdd\u5b58\u7684k<\/p>\n
![](\"https:\/\/cdn.myapp.ltd\/ag\/131\/sucai\/21.jpg\")
<\/figure>\n<\/p>\neras_model.h5\u6587\u4ef6\u653e\u5728\u540c\u4e00\u76ee\u5f55\u4e0b\uff0c\u5e76\u5c06\u8981\u9884\u6d4b\u7684\u624b\u5199\u6570\u5b57\u56fe\u7247\u547d\u540d\u4e3a`input_image.png`\uff0c\u7136\u540e\u8fd0\u884cexe\u6587\u4ef6\u3002\u8fd9\u6837\uff0c\u60a8\u5c31\u53ef\u4ee5\u5728\u4e0d\u9700\u8981Python\u73af\u5883\u7684\u8ba1\u7b97\u673a\u4e0a\u6267\u884cKeras\u6a21\u578b\u4e86\u3002<\/p>\n
\u6ce8\u610f\uff1a\u751f\u6210\u7684exe\u6587\u4ef6\u53ef\u80fd\u4f1a\u53d8\u5f97\u76f8\u5f53\u5927\uff0c\u56e0\u4e3a\u5b83\u9700\u8981\u5305\u542b\u8bb8\u591a\u4f9d\u8d56\u5e93\u3002\u4f46\u662f\uff0c\u8fd9\u662f\u4e00\u4e2a\u5feb\u901f\u5165\u95e8\u7684\u6559\u7a0b\uff0c\u8bb2\u89e3\u4e86\u5982\u4f55\u5c06Keras\u6a21\u578b\u6253\u5305\u6210exe\u6587\u4ef6\u7684\u57fa\u672c\u539f\u7406\u3002\u5982\u8981\u4f18\u5316\u751f\u6210\u7684exe\u6587\u4ef6\u5927\u5c0f\u6216\u6027\u80fd\uff0c\u53ef\u80fd\u9700\u8981\u8fdb\u4e00\u6b65\u4e86\u89e3PyInstaller\u7684\u9ad8\u7ea7\u9009\u9879\u4ee5\u53ca\u4f7f\u7528\u5176\u4ed6\u5de5\u5177\u8fdb\u884c\u6253\u5305\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"
\u5728\u672c\u6559\u7a0b\u4e2d\uff0c\u6211\u4eec\u5c06\u5b66\u4e60\u5982\u4f55\u5c06Keras\u6a21\u578b\u6253\u5305\u6210\u4e00\u4e2a\u72ec\u7acb\u7684exe\u6587\u4ef6\uff0c\u4ee5\u4fbf\u60a8\u53ef\u4ee5\u5728\u6ca1\u6709Python\u73af\u5883\u7684\u8ba1\u7b97\u673a\u4e0a\u8fd0\u884c\u5b83\u3002\u4f7f\u7528Keras\u4f5c\u4e3a\u6df1\u5ea6\u5b66\u4e60\u6846\u67b6\uff0c\u4f7f\u7528pyinstaller\u6765\u5c06\u811a\u672c\u6253\u5305\u6210\u4e00\u4e2a\u72ec<\/p>\n","protected":false},"author":4,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[3],"tags":[32514,32513,12,6509,2306],"topic":[],"class_list":["post-43631","post","type-post","status-publish","format-standard","hentry","category-zhuomianruanjian","tag-setupexe","tag-csharpexe","tag-12","tag-6509","tag-2306"],"_links":{"self":[{"href":"https:\/\/www.zhidianwl.net\/zhidianwl\/wp-json\/wp\/v2\/posts\/43631","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.zhidianwl.net\/zhidianwl\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.zhidianwl.net\/zhidianwl\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.zhidianwl.net\/zhidianwl\/wp-json\/wp\/v2\/users\/4"}],"replies":[{"embeddable":true,"href":"https:\/\/www.zhidianwl.net\/zhidianwl\/wp-json\/wp\/v2\/comments?post=43631"}],"version-history":[{"count":1,"href":"https:\/\/www.zhidianwl.net\/zhidianwl\/wp-json\/wp\/v2\/posts\/43631\/revisions"}],"predecessor-version":[{"id":43655,"href":"https:\/\/www.zhidianwl.net\/zhidianwl\/wp-json\/wp\/v2\/posts\/43631\/revisions\/43655"}],"wp:attachment":[{"href":"https:\/\/www.zhidianwl.net\/zhidianwl\/wp-json\/wp\/v2\/media?parent=43631"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.zhidianwl.net\/zhidianwl\/wp-json\/wp\/v2\/categories?post=43631"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.zhidianwl.net\/zhidianwl\/wp-json\/wp\/v2\/tags?post=43631"},{"taxonomy":"topic","embeddable":true,"href":"https:\/\/www.zhidianwl.net\/zhidianwl\/wp-json\/wp\/v2\/topic?post=43631"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}