keras merged model

最新推荐文章于 2020-06-08 15:20:14 发布

jacob6_b

最新推荐文章于 2020-06-08 15:20:14 发布

阅读量345

点赞数

分类专栏： python学习 keras

本文链接：https://blog.csdn.net/jacob6_b/article/details/78468617

版权

python学习同时被 2 个专栏收录

5 篇文章

订阅专栏

keras

3 篇文章

订阅专栏

本文介绍了一种使用Keras框架实现的手写数字识别系统。该系统通过合并三个不同分支的卷积神经网络来提高识别准确性。每个分支针对输入图像的不同特征进行训练，最终通过全连接层整合所有特征进行分类。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

keras merged model
可参考网址
http://www.cnblogs.com/qianboping/p/6509794.html
https://www.kaggle.com/nikosias/keras-merged-model

import numpy as np
import pandas as pd

from keras.optimizers import SGD
from keras.models import Sequential
from keras.layers import Merge
from keras.layers.convolutional import Convolution2D, MaxPooling2D, ZeroPadding2D
from keras.layers.core import Dense, Activation, Dropout, Reshape, Flatten
from keras.utils.np_utils import to_categorical

data = pd.read_csv('../input/train.csv')
images = data.iloc[:,1:].values
images = images.astype(np.float)
images = np.multiply(images, 1.0 / 255.0)
print ('Train\'s shape =>({0[0]},{0[1]})'.format(images.shape))

labelsFlat = data[[0]].values
classes = len(np.unique(labelsFlat))
labelsCategorical = to_categorical(labelsFlat,classes)
labelsCategorical = labelsCategorical.astype(np.uint8)
print ('Train\'s classes =>({0})'.format(classes))

test = pd.read_csv('../input/test.csv').values
testX = test
testX = testX.astype(np.float)
testX = np.multiply(testX, 1.0 / 255.0)
print ('Test\'s shape =>({0[0]},{0[1]})'.format(testX.shape))

leftBranch = Sequential()
leftBranch.add(Reshape((1,28,28), input_shape=(784,)))
leftBranch.add(Convolution2D(classes, 3, 1, activation='relu'))
leftBranch.add(MaxPooling2D((2, 2), strides=(2, 2)))
leftBranch.add(Flatten())

rightBranch = Sequential()
rightBranch.add(Reshape((1,28,28), input_shape=(784,)))
rightBranch.add(Convolution2D(classes, 1, 3, activation='relu'))
rightBranch.add(MaxPooling2D((2, 2), strides=(2, 2)))
rightBranch.add(Flatten())

centralBranch = Sequential()
centralBranch.add(Reshape((1,28,28), input_shape=(784,)))
centralBranch.add(Convolution2D(classes, 5, 5, activation='relu'))
centralBranch.add(MaxPooling2D((2, 2), strides=(2, 2)))
centralBranch.add(Flatten())

merged = Merge([leftBranch, centralBranch, rightBranch], mode='concat')

model = Sequential()
model.add(merged)
model.add(Dense(28*3, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(28, activation='relu'))
model.add(Dense(input_dim=10, output_dim=classes))
model.add(Activation("softmax"))

sgd = SGD(lr=0.5, momentum=0.0, decay=0.0, nesterov=False)
model.compile(loss='categorical_crossentropy', optimizer=sgd, metrics=['accuracy'])
model.fit([images,images,images], labelsCategorical,
            nb_epoch=5, batch_size=100, verbose=2)

yPred = model.predict_classes([testX,testX,testX])

np.savetxt('dr.csv', np.c_[range(1,len(yPred)+1),yPred], delimiter=',',
            header = 'ImageId,Label', comments = '', fmt='%d')