def model1():
numEpochs = 50
lrRate = 1e-2
lrRateDecay = lrRate/numEpochs
model = Model.build(width=48, height=48, depth=3, classes=2)
opt = Adagrad(lr=lrRate, decay=lrRateDecay)
model.compile(loss='binary_crossentropy',
optimizer=opt,
metrics=['accuracy'])
keras2ascii(model)
#plot_model(model, to_file='model.png')
return model
def plot_model(model, filename='model.png'):
from keras.utils import plot_model
plot_model(model,
to_file=filename,
show_layer_names=True,
show_shapes=True)
from keras_sequential_ascii import keras2ascii
print(keras2ascii(model))
model.add(MaxPooling2D(pool_size=(2, 2)))
# 对于池化层的输出，采用0.35概率的Dropout
model.add(Dropout(0.35))
# 展平所有像素，比如[28*28] -> [784]
model.add(Flatten())
# 对所有像素使用全连接层，输出为128，激活函数选用relu
model.add(Dense(128, activation='relu'))
# 对输入采用0.5概率的Dropout
model.add(Dropout(0.5))
# 对刚才Dropout的输出采用softmax激活函数，得到最后结果0-9
model.add(Dense(num_classes, activation='softmax'))
# 模型我们使用交叉熵损失函数，最优化方法选用Adadelta
model.compile(loss=keras.metrics.categorical_crossentropy,
optimizer=keras.optimizers.Adadelta(),
metrics=['accuracy'])
# 令人兴奋的训练过程
model.fit(x_train,
y_train,
batch_size=batch_size,
epochs=epochs,
verbose=1,
validation_data=(x_test, y_test))
score = model.evaluate(x_test, y_test, verbose=0)
model.save("model.h5")
keras2ascii(model) # 显示模型可视化
print('Test loss:', score[0])
print('Test accuracy:', score[1])
# EVALUATE MODELS
# In[110]:
score = model.evaluate(x_test, y_test, verbose=0)
# Print out accuracy. Change to score[0] for loss.
print "%s %.5f" % (f, score[1])
# Could add some code to predict on a new sample
# model.precict(new_sample)
# Print visual of model layers
#import keras_sequential_ascii as ksq
#ksq.sequential_model_to_ascii_printout(cnn_n)
from keras_sequential_ascii import keras2ascii
keras2ascii(model)
# Confusion Matrix
from sklearn.metrics import classification_report, confusion_matrix
Y_pred = model.predict(x_test, verbose=2)
y_pred = np.argmax(Y_pred, axis=1)
for ix in range(10):
print(ix,
confusion_matrix(np.argmax(y_test, axis=1), y_pred)[ix].sum())
cm = confusion_matrix(np.argmax(y_test, axis=1), y_pred)
print(cm)
# Visualizing of confusion matrix
import seaborn as sn
classificadorx2.add(Dropout(0.1))
classificadorx2.add(Dense(units = 128,
activation = 'relu',
use_bias=True,
kernel_initializer=initializers.RandomNormal(stddev=0.01),
bias_initializer='zeros'))
classificadorx2.add(Dropout(0.05))
classificadorx2.add(Dense(units = 18,
activation = 'sigmoid'))
classificadorx2.compile(optimizer = 'Adam',
loss = 'categorical_crossentropy',
metrics = ['accuracy'])
classificadorx2.summary()
keras2ascii(classificadorx2)
epochs = 500
learning_rate = ReduceLROnPlateau(monitor='accuracy',
factor=0.1,
patience=2,
verbose=1,
mode="auto",
min_delta=0.001,
cooldown=0,
min_lr=0.01)
h_x2 = classificadorx2.fit(base_treino,
steps_per_epoch = 100,
epochs = epochs,
def build_model(lr, batch_size, epochs, activation_function, optimizer,
conv_depth, model_name):
graph = K.get_session().graph
with graph.as_default():
adam = Adam(lr=lr, decay=1e-6)
sgd = SGD(lr=lr, decay=1e-6)
rmsprop = RMSprop(lr=lr, decay=1e-6)
optimizers = {
'adam': adam,
'sgd': sgd,
'rmsprop': rmsprop,
}
model = Sequential()
filters = 32
model.add(
Conv2D(filters,
kernel_size=(3, 3),
activation=activation_function,
input_shape=(32, 32, 3),
padding='same'))
# model.add(Conv2D(filters, kernel_size=(3, 3),
# activation=activation_function,
# padding='same'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(rate=0.25, seed=123))
filters *= 2
for _ in range(conv_depth - 1):
model.add(
Conv2D(filters,
kernel_size=(3, 3),
activation=activation_function,
padding='same'))
# model.add(Conv2D(filters, kernel_size=(3, 3),
# activation=activation_function,
# padding='same'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(rate=0.25, seed=123))
filters *= 2
model.add(Flatten())
for _ in range(1):
model.add(Dense(1024, activation=activation_function))
model.add(Dropout(rate=0.5, seed=123))
num_classes = 10
model.add(Dense(num_classes, activation='softmax'))
print(optimizers[optimizer])
model.compile(loss='categorical_crossentropy',
optimizer=adam,
metrics=['accuracy'])
keras2ascii(model)
X_train, Y_train, X_test, Y_test = load_data(num_classes)
history = model.fit(X_train,
Y_train,
batch_size=batch_size,
shuffle=True,
epochs=epochs,
validation_data=(X_test, Y_test),
verbose=1)
save_model(model, model_name)
make_plot(history, epochs, model_name)
