基于 Keras 搭建 CNN 模型

import numpy as np
import matplotlib.pyplot as plt
import keras
import random, csv
from keras import layers, models
from keras.models import Model
from keras.datasets import mnist
from keras.models import load_model, Sequential
from keras.preprocessing import image
from keras.layers import Dense, Dropout, Activation, Flatten, Conv2D
from keras.layers.convolutional import Convolution2D, MaxPooling2D
from keras.utils import np_utils, to_categorical
from tensorflow.examples.tutorials.mnist import input_data
from dnn_utils_v2 import plt_sample
from inputDataProc import intialize_data_TS

def xxCNN(train_x, train_y, train_y_orig, test_x, test_y, test_y_orig, label_interval):
	
	classes = len(label_interval)
	#plt_sample(train_y, test_y, label_interval)
	print(train_x.shape)
	print(test_x.shape)
	print(train_y.shape)
	print(test_y.shape)
	#one-hot编码
	train_y = np_utils.to_categorical(train_y, classes)
	test_y = np_utils.to_categorical(test_y, classes)
	#kernel_regularizer=keras.regularizers.l2(0.0001))
	#model = load_model('D://xx.h5')
	#'''
	model = Sequential()
	model.add(Convolution2D(1, kernel_size=(1, 3), strides=1, padding='valid', input_shape=(1, train_x.shape[2], train_x.shape[3]), data_format = "channels_first", kernel_regularizer=keras.regularizers.l2(0)))
	model.add(Activation('relu'))
	#model.add(Dropout(0.1))
	#model.add(Conv2D(16, kernel_size=(2, 2), strides=2, activation='relu', kernel_regularizer=keras.regularizers.l2(0.001)))
	#model.add(MaxPooling2D(pool_size=(2, 2)))
	#model.add(Dropout(0.1))
	#model.add(Conv2D(16, kernel_size=(2, 2), strides=2, activation='relu', kernel_regularizer=keras.regularizers.l2(0.001)))
	#model.add(Dropout(0.1))
	#model.add(Conv2D(16, (2, 1), activation='relu'))
	
	model.add(Flatten())
	model.add(Dense(64, kernel_regularizer=keras.regularizers.l2(0.001)))
	model.add(Activation('relu'))
	model.add(Dropout(0.1))
	model.add(Dense(classes))                    #类别个数
	model.add(Activation('softmax'))

	model.compile(loss='categorical_crossentropy', optimizer='sgd', metrics=['accuracy'])
	
	model.summary()
	model.get_config()
	#reduce_lr = keras.callbacks.ReduceLROnPlateau(monitor='val_loss', patience=1, mode='auto')
	history = model.fit(train_x, train_y, batch_size=256, epochs=50, verbose=1, validation_data=(test_x, test_y))#, callbacks=[reduce_lr])
	score = model.evaluate(test_x, test_y, verbose=1)
	#模型保存JSON文件
	model_json = model.to_json()
	with open("D://xx.json", 'w') as file:
		file.write(model_json)
	#保存模型权重值
	model.save('D://xx.h5')
	#'''
	#'''
	test_xx = test_x[0]
	test_x_temp = test_xx.reshape([10, 37])
	
	#取一个样本写入csv文件，测试c++代码使用
	with open("D://xx.csv", "w", newline='') as csvfile: 
		writer = csv.writer(csvfile)
		#写入多行用writerows
		writer.writerows(test_x_temp)
		csvfile.close()
	
	test_xx = test_xx.reshape([1, 1, 10, 37])
	conv2d_1_model = Model(inputs = model.input, outputs = model.get_layer('conv2d_1').output)
	conv2d_1_res = conv2d_1_model.predict(test_xx)
	dense_1_model = Model(inputs = model.input, outputs = model.get_layer('dense_1').output)
	dense_1_res = dense_1_model.predict(test_xx)
	dense_2_model = Model(inputs = model.input, outputs = model.get_layer('dense_2').output)
	dense_2_res = dense_2_model.predict(test_xx)
	#'''
	result = model.predict(test_xx)
	print(result)
	result = np.argmax(result, axis = 1)
	print(result)
	
	out_res(result, test_y_orig, label_interval)
	print(result)
	print('Test score:', score[0])
	print('Test accuracy:', score[1])

	score = model.evaluate(train_x, train_y, verbose=1)
	result = model.predict(train_x)
	result = np.argmax(result, axis=1)

	out_res(result, train_y_orig, label_interval)
	print(result)
	print('Test score:', score[0])
	print('Test accuracy:', score[1])