python CNN
- 教育綜合
- 2023-01-27 17:43:25
Python keras構(gòu)建CNN
data.py:
#coding:utf-8
"""
Author:wepon
Source:https://github.com/wepe
"""
importos
fromPILimportImage
importnumpyasnp
#讀取文件夾mnist下的42000張圖片，圖片為灰度圖，所以為1通道，圖像大小28*28
#如果是將彩色圖作為輸入,則將1替換為3，并且data[i,:,:,:]=arr改為data[i,:,:,:]=[arr[:,:,0],arr[:,:,1],arr[:,:,2]]
defload_data():
data=np.empty((42000,1,28,28),dtype="float32")
label=np.empty((42000,),dtype="uint8")
imgs=os.listdir("./mnist")
num=len(imgs)
foriinrange(num):
img=Image.open("./mnist/"+imgs[i])
arr=np.asarray(img,dtype="float32")
data[i,:,:,:]=arr
label[i]=int(imgs[i].split('.')[0])
returndata,label
由于Keras系統(tǒng)升級，cnn.py代碼調(diào)整如下：
#coding:utf-8
'''
GPUruncommand:
THEANO_FLAGS=mode=FAST_RUN,device=gpu,floatX=float32pythoncnn.py
CPUruncommand:
pythoncnn.py
'''
#導入各種用到的模塊組件
from__future__importabsolute_import
from__future__importprint_function
fromkeras.preprocessing.imageimportImageDataGenerator
fromkeras.modelsimportSequential
fromkeras.layers.coreimportDense,Dropout,Activation,Flatten
fromkeras.layers.advanced_activationsimportPReLU
fromkeras.layers.convolutionalimportConvolution2D,MaxPooling2D
fromkeras.optimizersimportSGD,Adadelta,Adagrad
fromkeras.utilsimportnp_utils,generic_utils
fromsix.movesimportrange
fromdataimportload_data
importrandom
#加載數(shù)據(jù)
data,label=load_data()
#打亂數(shù)據(jù)
index=[iforiinrange(len(data))]
random.shuffle(index)
data=data[index]
label=label[index]
print(data.shape[0],'samples')
#label為0~9共10個類別，keras要求格式為binaryclassmatrices,轉(zhuǎn)化一下，直接調(diào)用keras提供的這個函數(shù)
label=np_utils.to_categorical(label,10)怎樣用python構(gòu)建一個卷積神經(jīng)網(wǎng)絡
用keras框架較為方便
首先安裝anaconda，然后通過pip安裝keras
以下轉(zhuǎn)自wphh的博客。
#coding:utf-8
'''
GPUruncommand:
THEANO_FLAGS=mode=FAST_RUN,device=gpu,floatX=float32pythoncnn.py
CPUruncommand:
pythoncnn.py
2016.06.06更新：
這份代碼是keras開發(fā)初期寫的，當時keras還沒有現(xiàn)在這么流行，文檔也還沒那么豐富，所以我當時寫了一些簡單的教程。
現(xiàn)在keras的API也發(fā)生了一些的變化，建議及推薦直接上keras.io看更加詳細的教程。
'''
#導入各種用到的模塊組件
from__future__importabsolute_import
from__future__importprint_function
fromkeras.preprocessing.imageimportImageDataGenerator
fromkeras.modelsimportSequential
fromkeras.layers.coreimportDense,Dropout,Activation,Flatten
fromkeras.layers.advanced_activationsimportPReLU
fromkeras.layers.convolutionalimportConvolution2D,MaxPooling2D
fromkeras.optimizersimportSGD,Adadelta,Adagrad
fromkeras.utilsimportnp_utils,generic_utils
fromsix.movesimportrange
fromdataimportload_data
importrandom
importnumpyasnp
np.random.seed(1024)#forreproducibility
#加載數(shù)據(jù)
data,label=load_data()
#打亂數(shù)據(jù)
index=[iforiinrange(len(data))]
random.shuffle(index)
data=data[index]
label=label[index]
print(data.shape[0],'samples')
#label為0~9共10個類別，keras要求格式為binaryclassmatrices,轉(zhuǎn)化一下，直接調(diào)用keras提供的這個函數(shù)
label=np_utils.to_categorical(label,10)如何用caffe的python接口實現(xiàn)cnn
簡單的說，你需要把py-faster-rcnn下的caffe-fast-rcnn遷移到Win下重新編譯，主要是為了編譯pycaffe，開啟WITH_PYTHON_LAYER，還要在Win下把lib目錄下的python代碼重新編譯以上兩項我都用的是CPU編譯這樣就可以運行Tools下的Demo.py了，參數(shù)--cpu。模型文件我用linux下訓練好的。怎樣用python構(gòu)建一個卷積神經(jīng)網(wǎng)絡模型
上周末利用python簡單實現(xiàn)了一個卷積神經(jīng)網(wǎng)絡，只包含一個卷積層和一個maxpooling層，pooling層后面的多層神經(jīng)網(wǎng)絡采用了softmax形式的輸出。實驗輸入仍然采用MNIST圖像使用10個feature map時，卷積和pooling的結(jié)果分別如下所示。
部分源碼如下：
[python]view plaincopy
#coding=utf-8
'''''
Createdon2014年11月30日
@author:Wangliaofan
'''
importnumpy
importstruct
importmatplotlib.pyplotasplt
importmath
importrandom
importcopy
#test
fromBasicMultilayerNeuralNetworkimportBMNN2
defsigmoid(inX):
if1.0+numpy.exp(-inX)==0.0:
return999999999.999999999
return1.0/(1.0+numpy.exp(-inX))
defdifsigmoid(inX):
returnsigmoid(inX)*(1.0-sigmoid(inX))
deftangenth(inX):
return(1.0*math.exp(inX)-1.0*math.exp(-inX))/(1.0*math.exp(inX)+1.0*math.exp(-inX))
defcnn_conv(in_image,filter_map,B,type_func='sigmoid'):
#in_image[num,featuremap,row,col]=>in_image[Irow,Icol]
#featuresmap[kfilter,row,col]
#type_func['sigmoid','tangenth']
#out_feature[kfilter,Irow-row+1,Icol-col+1]
shape_image=numpy.shape(in_image)#[row,col]
#print"shape_image",shape_image
shape_filter=numpy.shape(filter_map)#[kfilter,row,col]
ifshape_filter[1]>shape_image[0]orshape_filter[2]>shape_image[1]:
raiseException
shape_out=(shape_filter[0],shape_image[0]-shape_filter[1]+1,shape_image[1]-shape_filter[2]+1)
out_feature=numpy.zeros(shape_out)
k,m,n=numpy.shape(out_feature)
fork_idxinrange(0,k):
#rotate180tocalculateconv
c_filter=numpy.rot90(filter_map[k_idx,:,:],2)
forr_idxinrange(0,m):
forc_idxinrange(0,n):
#conv_temp=numpy.zeros((shape_filter[1],shape_filter[2]))
conv_temp=numpy.dot(in_image[r_idx:r_idx+shape_filter[1],c_idx:c_idx+shape_filter[2]],c_filter)
sum_temp=numpy.sum(conv_temp)
iftype_func=='sigmoid':
out_feature[k_idx,r_idx,c_idx]=sigmoid(sum_temp+B[k_idx])
eliftype_func=='tangenth':
out_feature[k_idx,r_idx,c_idx]=tangenth(sum_temp+B[k_idx])
else:
raiseException
returnout_feature
defcnn_maxpooling(out_feature,pooling_size=2,type_pooling="max"):
k,row,col=numpy.shape(out_feature)
max_index_Matirx=numpy.zeros((k,row,col))
out_row=int(numpy.floor(row/pooling_size))
out_col=int(numpy.floor(col/pooling_size))
out_pooling=numpy.zeros((k,out_row,out_col))
fork_idxinrange(0,k):
forr_idxinrange(0,out_row):
forc_idxinrange(0,out_col):
temp_matrix=out_feature[k_idx,pooling_size*r_idx:pooling_size*r_idx+pooling_size,pooling_size*c_idx:pooling_size*c_idx+pooling_size]
out_pooling[k_idx,r_idx,c_idx]=numpy.amax(temp_matrix)
max_index=numpy.argmax(temp_matrix)
#printmax_index
#printmax_index/pooling_size,max_index%pooling_size
max_index_Matirx[k_idx,pooling_size*r_idx+max_index/pooling_size,pooling_size*c_idx+max_index%pooling_size]=1
returnout_pooling,max_index_Matirx
defpoolwithfunc(in_pooling,W,B,type_func='sigmoid'):
k,row,col=numpy.shape(in_pooling)
out_pooling=numpy.zeros((k,row,col))
fork_idxinrange(0,k):
forr_idxinrange(0,row):
forc_idxinrange(0,col):
out_pooling[k_idx,r_idx,c_idx]=sigmoid(W[k_idx]*in_pooling[k_idx,r_idx,c_idx]+B[k_idx])
returnout_pooling
#out_featureistheoutputofconv
defbackErrorfromPoolToConv(theta,max_index_Matirx,out_feature,pooling_size=2):
k1,row,col=numpy.shape(out_feature)
error_conv=numpy.zeros((k1,row,col))
k2,theta_row,theta_col=numpy.shape(theta)
ifk1!=k2:
raiseException
foridx_kinrange(0,k1):
foridx_rowinrange(0,row):
foridx_colinrange(0,col):
error_conv[idx_k,idx_row,idx_col]=\
max_index_Matirx[idx_k,idx_row,idx_col]*\
float(theta[idx_k,idx_row/pooling_size,idx_col/pooling_size])*\
difsigmoid(out_feature[idx_k,idx_row,idx_col])
returnerror_conv
defbackErrorfromConvToInput(theta,inputImage):
k1,row,col=numpy.shape(theta)
#print"theta",k1,row,col
i_row,i_col=numpy.shape(inputImage)
ifrow>i_roworcol>i_col:
raiseException
filter_row=i_row-row+1
filter_col=i_col-col+1
detaW=numpy.zeros((k1,filter_row,filter_col))
#thesamewithconvvalidinmatlab
fork_idxinrange(0,k1):
foridx_rowinrange(0,filter_row):
foridx_colinrange(0,filter_col):
subInputMatrix=inputImage[idx_row:idx_row+row,idx_col:idx_col+col]
#print"subInputMatrix",numpy.shape(subInputMatrix)
#rotatetheta180
#printnumpy.shape(theta)
theta_rotate=numpy.rot90(theta[k_idx,:,:],2)
#print"theta_rotate",theta_rotate
dotMatrix=numpy.dot(subInputMatrix,theta_rotate)
detaW[k_idx,idx_row,idx_col]=numpy.sum(dotMatrix)
detaB=numpy.zeros((k1,1))
fork_idxinrange(0,k1):
detaB[k_idx]=numpy.sum(theta[k_idx,:,:])
returndetaW,detaB
defloadMNISTimage(absFilePathandName,datanum=60000):
images=open(absFilePathandName,'rb')
buf=images.read()
index=0
magic,numImages,numRows,numColumns=struct.unpack_from('>IIII',buf,index)
printmagic,numImages,numRows,numColumns
index+=struct.calcsize('>IIII')
ifmagic!=2051:
raiseException
datasize=int(784*datanum)
datablock=">"+str(datasize)+"B"
#nextmatrix=struct.unpack_from('>47040000B',buf,index)
nextmatrix=struct.unpack_from(datablock,buf,index)
nextmatrix=numpy.array(nextmatrix)/255.0
#nextmatrix=nextmatrix.reshape(numImages,numRows,numColumns)
#nextmatrix=nextmatrix.reshape(datanum,1,numRows*numColumns)
nextmatrix=nextmatrix.reshape(datanum,1,numRows,numColumns)
returnnextmatrix,numImages
defloadMNISTlabels(absFilePathandName,datanum=60000):
labels=open(absFilePathandName,'rb')
buf=labels.read()
index=0
magic,numLabels=struct.unpack_from('>II',buf,index)
printmagic,numLabels
index+=struct.calcsize('>II')
ifmagic!=2049:
raiseException
datablock=">"+str(datanum)+"B"
#nextmatrix=struct.unpack_from('>60000B',buf,index)
nextmatrix=struct.unpack_from(datablock,buf,index)
nextmatrix=numpy.array(nextmatrix)
returnnextmatrix,numLabels
defsimpleCNN(numofFilter,filter_size,pooling_size=2,maxIter=1000,imageNum=500):
decayRate=0.01
MNISTimage,num1=loadMNISTimage("F:\MachineLearning\UFLDL\data\common\\train-images-idx3-ubyte",imageNum)
printnum1
row,col=numpy.shape(MNISTimage[0,0,:,:])
out_Di=numofFilter*((row-filter_size+1)/pooling_size)*((col-filter_size+1)/pooling_size)
MLP=BMNN2.MuiltilayerANN(1,[128],out_Di,10,maxIter)
MLP.setTrainDataNum(imageNum)
MLP.loadtrainlabel("F:\MachineLearning\UFLDL\data\common\\train-labels-idx1-ubyte")
MLP.initialweights()
#MLP.printWeightMatrix()
rng=numpy.random.RandomState(23455)
W_shp=(numofFilter,filter_size,filter_size)
W_bound=numpy.sqrt(numofFilter*filter_size*filter_size)
W_k=rng.uniform(low=-1.0/W_bound,high=1.0/W_bound,size=W_shp)
B_shp=(numofFilter,)
B=numpy.asarray(rng.uniform(low=-.5,high=.5,size=B_shp))
cIter=0
whilecIter
cIter+=1
ImageNum=random.randint(0,imageNum-1)
conv_out_map=cnn_conv(MNISTimage[ImageNum,0,:,:],W_k,B,"sigmoid")
out_pooling,max_index_Matrix=cnn_maxpooling(conv_out_map,2,"max")
pool_shape=numpy.shape(out_pooling)
MLP_input=out_pooling.reshape(1,1,out_Di)
#printnumpy.shape(MLP_input)
DetaW,DetaB,temperror=MLP.backwardPropogation(MLP_input,ImageNum)
ifcIter%50==0:
printcIter,"Temperror:",temperror
#printnumpy.shape(MLP.Theta[MLP.Nl-2])
#printnumpy.shape(MLP.Ztemp[0])
#printnumpy.shape(MLP.weightMatrix[0])
theta_pool=MLP.Theta[MLP.Nl-2]*MLP.weightMatrix[0].transpose()
#printnumpy.shape(theta_pool)
#print"theta_pool",theta_pool
temp=numpy.zeros((1,1,out_Di))
temp[0,:,:]=theta_pool
back_theta_pool=temp.reshape(pool_shape)
#print"back_theta_pool",numpy.shape(back_theta_pool)
#print"back_theta_pool",back_theta_pool
error_conv=backErrorfromPoolToConv(back_theta_pool,max_index_Matrix,conv_out_map,2)
#print"error_conv",numpy.shape(error_conv)
#printerror_conv
conv_DetaW,conv_DetaB=backErrorfromConvToInput(error_conv,MNISTimage[ImageNum,0,:,:])
#print"W_k",W_k
#print"conv_DetaW",conv_DetaW
python怎么將預訓練好的詞向量初始化cnn
因為你從stopword.txt這個文件里讀出來的是是str類型的，str類型跟unicode不相等
下一篇
返回列表
