Deep learning Python Script Training Keras mnist digital recognition model __python

This script is a training Keras mnist digital Recognition program, previously sent, today to achieve the forecast,

# larger CNN for the mnist Dataset # 2.Negative dimension size caused by subtracting 5 from 1 for ' conv2d_4/convolution ' ( OP: ' conv2d ') with input shapes # 3.userwarning:update your ' conv2d ' call to the Keras 2 Api:http://blog.csdn.net/johini eli/article/details/69222956 # 4.Error checking input:expected conv2d_1_input to have shape (None, a, 1) but go

T array with shape (60000, 1, a) # talk to Wumi,you good. # python 3.5.4 # keras.__version__: 2.0.6 # thensorflow 1.2.1 # Theano 0.10.0beta1 # Good Blog # http://blog.csdn.net/s hizhengxin123/article/details/72383728 # http://www.360doc.com/content/17/0415/12/1489589_645772879.shtml #  Recommand Another framework http://tflearn.org/examples/import numpy import Keras from keras.datasets import mnist from Keras.models import sequential from keras.layers import dense to keras.layers import dropout from keras.layers import F Latten from keras.layers.convolutional import conv2d from keras.layers.convolutional import MAxpooling2d from keras.utils import np_utils import Matplotlib.pyplot as plt to keras.constraints import maxnorm from Ke 
Ras.optimizers Import SGD from keras.preprocessing import image import Skimage.io # fix random seed for reproducibility

Seed = 7 Numpy.random.seed (SEED) # Load data (X_train, Y_train), (x_test, y_test) = Mnist.load_data () plt.subplot (221) Plt.imshow (X_train[0], Cmap=plt.get_cmap (' Gray ')) Plt.show () # Reshape to be [samples][pixels][width][height] X_train = X_train.reshape (X_train.shape[0], 1). Astype (' float32 ') x_test = X_test.reshape (x_test.shape[0), 28, 28, 1). Astype (' float32 ') # X_train = X_train.reshape (1, 1). Astype (' float32 ') valueerror:cannot reshape array of size 470  40000 into shape (1,28,28,1) #X_test = X_test.reshape (1, 1). Astype (' float32 ') valueerror:cannot reshape array of Size 47040000 into shape (1,28,28,1) # X_train = X_train.reshape (X_train.shape[0], 1, a). Astype (' float32 ') # X_test = X_test.reshape (X_test.shape[0]1, Astype (' float32 ') <---4 # normalize inputs from 0-255 to 0-1 x_train = x_train/255 x_test = x_test/ 255 # One hot encode outputs Y_train = np_utils.to_categorical (y_train) y_test = np_utils.to_categorical (y_test) Num_class
    es = y_test.shape[1] # # #raw # define the larger model Def Larger_model (): # Create Model model = sequential () Model.add (conv2d (5, 5), padding= ' valid ', input_shape= (1), activation= ' Relu ')) # Model.add (30, 5 , 5), padding= ' valid ', input_shape= (28,1), activation= ' Relu ')) <----3,2 Model.add (maxpooling2d (2, 2)) Model.add (Dropout (0.4)) Model.add (conv2d (3, 3), activation= ' Relu ') Model.add (Maxpooling2d (pool_size = (2, 2)) Model.add (Dropout (0.4)) Model.add (Flatten ()) Model.add (dense (128, activation= ' Relu ')) Model.add ( Dropout (0.4)) Model.add (Dense (activation= ' relu ')) Model.add (Dropout (0.4)) Model.add (Dense (num_classes, act
    ivation= ' Softmax ')# Compile Model # Optimizer Optimizer # loss loss function Model.compile (loss= ' categorical_crossentropy ', optimizer= ' Adam ', metrics=[' accuracy '] return model # builds the model = Larger_model () # Fit the Model # Fit function returns a History object whose h The Istory.history property records the change of the value of the loss function and other metrics with the epoch, and if there is a validation set, it also contains the changes of these indices of the validation set Model.fit (X_train, Y_train, validation_data= (X_test, Y_test), epochs=10, batch_size=200, verbose=2) # epochs too bigger # Model.fit (X_train, Y_train, V Alidation_data= (X_test, Y_test), nb_epoch=200, batch_size=200, verbose=2) # Final evaluation of the model scores = MODEL.E Valuate (X_test, Y_test, verbose=0) print ("Large CNN Error:%.2f%%"% (100-scores[1] *)) # Save the Model Model.save (' D:\\works\\jetbrians\\pycharmprojects\\trypicture\\my_model.h5 ') # Creates a HDF5 file ' My_model.h5 ' del model # Reloa d The Modle # returns a compiled model # identical to the previous one # modeltrained = sequential () # model = Modeltraine D.load_model (' D:\\works\\jetbrianS\\pycharmprojects\\trypicture\\my_model.h5 ') # https://gist.github.com/ageitgey/  A40DDED08E82E59724C70DA23786BBF0 # Write a number in a picture # predict numbers #image_path = './lena.jpg ' # method 1 # Load pic #img = image.load_img (Image_path, target_size=) # handle pic #x = Image.img_to_array (img) #x = numpy.exp And_dims (x, axis=0) #x = Preprocess_input (x) # Method2 #img2 = Skimage.io.imread (Image_path, As_grey=true) #skimage. io.im Show (IMG2) #plt. Show () #img2 = Numpy.reshape (Img2, (1, 1)). Astype (' float32 ') # predicts numbers #https://baijiahao.baidu.c OM/S?ID=1574962680356106&AMP;WFR=SPIDER&AMP;FOR=PC #predict = model.predict (Img2, verbose=0) #result = Model.prediect  _classes (Img2, verbose=0) #print (predict[0]) #print (result[0)) #some warning tips The TensorFlow library wasn ' t compiled
To use AVX2 instructions, but this are available on your machine and could the up CPU speed. #have no idea what ' s the meaning original data style ================= training log D:\applications\ANaconda3\python.exe d:/works/jetbrians/pycharmprojects/trypicture/trainmodel/trainmodel.py Using TensorFlow
Backend. Train on 60000 samples, validate on 10000 samples Epoch 1/10 62S-LOSS:0.8830-ACC:0.7027-VAL_LOSS:0.1566-VAL_ACC : 0.9545 Epoch 2/10 56s-loss:0.3130-acc:0.9078-val_loss:0.0955-val_acc:0.9712 Epoch 3/10 61s-loss:0.2342-  acc:0.9340-val_loss:0.0737-val_acc:0.9763 Epoch 4/10 58S-LOSS:0.1924-ACC:0.9458-VAL_LOSS:0.0643-VAL_ACC: 0.9802 Epoch 5/10 60s-loss:0.1678-acc:0.9534-val_loss:0.0541-val_acc:0.9848 Epoch 6/10 cc:0.9578-val_loss:0.0468-val_acc:0.9849 Epoch 7/10 53S-LOSS:0.1396-ACC:0.9617-VAL_LOSS:0.0464-VAL_ACC: 0.9852 Epoch 8/10 55s-loss:0.1303-acc:0.9647-val_loss:0.0422-val_acc:0.9871 Epoch 9/10 c:0.9656-val_loss:0.0398-val_acc:0.9871 Epoch 10/10 53S-LOSS:0.1156-ACC:0.9680-VAL_LOSS:0.0370-VAL_ACC: 0.9876 Large CNN error:1.24%

Process finished with exit code 0