def initial_num_char_phase1():
"""
识别二值化图像的字符个数
:param bw: 二值图像
:return:
"""
# 加载模型
model = Sequential()
model.add(Convolution2D(4, 5, 5, input_shape=(1, 30, 40), border_mode='valid'))
model.add(Activation('tanh'))
model.add(Convolution2D(8, 5, 5, input_shape=(1, 26, 36), border_mode='valid'))
model.add(Activation('tanh'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.55))
model.add(Convolution2D(16, 4, 4, input_shape=(1, 11, 16), border_mode='valid'))
model.add(Activation('tanh'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.60))
model.add(Flatten())
model.add(Dense(input_dim=16*4*6, output_dim=256, init='glorot_uniform'))
model.add(Activation('tanh'))
model.add(Dense(input_dim=256, output_dim=2, init='glorot_uniform'))
model.add(Activation('softmax'))
# 加载权值
model.load_weights('model/train_len_size1.d5')
sgd = SGD(l2=0.0, lr=0.05, decay=1e-6, momentum=0.9, nesterov=True)
model.compile(loss='categorical_crossentropy', optimizer=sgd, class_mode="categorical")
return model
def test_nested_sequential(in_tmpdir):
(x_train, y_train), (x_test, y_test) = _get_test_data()
inner = Sequential()
inner.add(Dense(num_hidden, input_shape=(input_dim,)))
inner.add(Activation('relu'))
inner.add(Dense(num_class))
middle = Sequential()
middle.add(inner)
model = Sequential()
model.add(middle)
model.add(Activation('softmax'))
model.compile(loss='categorical_crossentropy', optimizer='rmsprop')
model.fit(x_train, y_train, batch_size=batch_size, epochs=epochs, verbose=1, validation_data=(x_test, y_test))
model.fit(x_train, y_train, batch_size=batch_size, epochs=epochs, verbose=2, validation_split=0.1)
model.fit(x_train, y_train, batch_size=batch_size, epochs=epochs, verbose=0)
model.fit(x_train, y_train, batch_size=batch_size, epochs=epochs, verbose=1, shuffle=False)
model.train_on_batch(x_train[:32], y_train[:32])
loss = model.evaluate(x_test, y_test, verbose=0)
model.predict(x_test, verbose=0)
model.predict_classes(x_test, verbose=0)
model.predict_proba(x_test, verbose=0)
fname = 'test_nested_sequential_temp.h5'
model.save_weights(fname, overwrite=True)
inner = Sequential()
inner.add(Dense(num_hidden, input_shape=(input_dim,)))
inner.add(Activation('relu'))
inner.add(Dense(num_class))
middle = Sequential()
middle.add(inner)
model = Sequential()
model.add(middle)
model.add(Activation('softmax'))
model.compile(loss='categorical_crossentropy', optimizer='rmsprop')
model.load_weights(fname)
os.remove(fname)
nloss = model.evaluate(x_test, y_test, verbose=0)
assert(loss == nloss)
# test serialization
config = model.get_config()
Sequential.from_config(config)
model.summary()
json_str = model.to_json()
model_from_json(json_str)
yaml_str = model.to_yaml()
model_from_yaml(yaml_str)
def test_merge_overlap():
left = Sequential()
left.add(Dense(nb_hidden, input_shape=(input_dim,)))
left.add(Activation('relu'))
model = Sequential()
model.add(Merge([left, left], mode='sum'))
model.add(Dense(nb_class))
model.add(Activation('softmax'))
model.compile(loss='categorical_crossentropy', optimizer='rmsprop')
model.fit(X_train, y_train, batch_size=batch_size, nb_epoch=nb_epoch, show_accuracy=True, verbose=1, validation_data=(X_test, y_test))
model.fit(X_train, y_train, batch_size=batch_size, nb_epoch=nb_epoch, show_accuracy=False, verbose=2, validation_data=(X_test, y_test))
model.fit(X_train, y_train, batch_size=batch_size, nb_epoch=nb_epoch, show_accuracy=True, verbose=2, validation_split=0.1)
model.fit(X_train, y_train, batch_size=batch_size, nb_epoch=nb_epoch, show_accuracy=False, verbose=1, validation_split=0.1)
model.fit(X_train, y_train, batch_size=batch_size, nb_epoch=nb_epoch, verbose=0)
model.fit(X_train, y_train, batch_size=batch_size, nb_epoch=nb_epoch, verbose=1, shuffle=False)
model.train_on_batch(X_train[:32], y_train[:32])
loss = model.evaluate(X_train, y_train, verbose=0)
assert(loss < 0.7)
model.predict(X_test, verbose=0)
model.predict_classes(X_test, verbose=0)
model.predict_proba(X_test, verbose=0)
model.get_config(verbose=0)
fname = 'test_merge_overlap_temp.h5'
model.save_weights(fname, overwrite=True)
model.load_weights(fname)
os.remove(fname)
nloss = model.evaluate(X_train, y_train, verbose=0)
assert(loss == nloss)
def getmodel():
nb_classes = 2
model = Sequential()
model.add(Convolution2D(32, 3, 3, border_mode='same',
input_shape=(1, RECEP_HEI, RECEP_WEI)))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Convolution2D(32, 3, 3))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Flatten())
model.add(Dense(128))
model.add(Activation('relu'))
model.add(Dropout(0.5))
model.add(Dense(nb_classes))
model.add(Activation('softmax'))
sgd = SGD(lr=0.1, decay=1e-6, momentum=0.9, nesterov=True)
model.load_weights('weights.hdf5')
model.compile(loss='categorical_crossentropy',
optimizer=sgd,
metrics=['accuracy'])
# model.fit(X_train, Y_train, batch_size=32, nb_epoch=1,
# verbose=1, shuffle = True ,validation_split=0.25)
return model
def neural_net(num_sensors, params, load=''):
model = Sequential()
# First layer.
model.add(Dense(
params[0], init='lecun_uniform', input_shape=(num_sensors,)
))
model.add(Activation('relu'))
model.add(Dropout(0.2))
# Second layer.
model.add(Dense(params[1], init='lecun_uniform'))
model.add(Activation('relu'))
model.add(Dropout(0.2))
# Output layer.
model.add(Dense(3, init='lecun_uniform'))
model.add(Activation('linear'))
rms = RMSprop()
model.compile(loss='mse', optimizer=rms)
if load:
model.load_weights(load)
return model
def get_model(self):
classes = 36
# data = np.empty((57218, 1, 24, 24), dtype="float32")
model = Sequential()
model.add(Convolution2D(4, 5, 5, border_mode='valid', input_shape=(1, 24, 24)))
model.add(BatchNormalization())
model.add(Activation('relu'))
model.add(Convolution2D(8, 3, 3, border_mode='valid'))
model.add(BatchNormalization())
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Convolution2D(16, 3, 3, border_mode='valid'))
model.add(BatchNormalization())
model.add(Activation('relu'))
# model.add(Dropout(0.5))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Flatten())
model.add(Dense(128, init='normal'))
model.add(BatchNormalization())
model.add(Activation('tanh'))
model.add(Dense(classes, init='normal'))
model.add(Activation('softmax'))
sgd = SGD(l2=0.0, lr=0.05, decay=1e-6, momentum=0.9, nesterov=True)
model.compile(loss='categorical_crossentropy', optimizer=sgd, class_mode="categorical")
model.load_weights("eduLogin/captcha/tmp/weights.11-0.05.h5")
return model
def inference_dense(input_dim, class_num, optimizer='sgd', weights_file=''):
model = Sequential()
model.add(Dense(2048, input_dim=input_dim))
model.add(Activation('relu'))
model.add(Dropout(0.5))
model.add(Dense(1024))
model.add(Activation('relu'))
model.add(Dropout(0.5))
# model.add(Dense(256))
# model.add(Activation('relu'))
# model.add(Dropout(0.5))
model.add(Dense(class_num))
model.add(Activation('softmax'))
if weights_file:
model.load_weights(weights_file)
# adadelta = Adadelta(lr=1.0, rho=0.95, epsilon=1e-06)
if optimizer == 'sgd':
opt = SGD(lr=1e-4, decay=1e-6, momentum=0.9, nesterov=True)
elif optimizer == 'adam':
opt = Adam(lr=0.0001, beta_1=0.9, beta_2=0.999, epsilon=1e-08)
elif optimizer == 'adagrad':
opt = Adagrad(lr=0.01, epsilon=1e-08)
elif optimizer == 'adadelta':
opt = Adadelta(lr=1.0, rho=0.95, epsilon=1e-08)
elif optimizer == 'rmsprop':
opt = RMSprop(lr=0.001, rho=0.9, epsilon=1e-08)
print('compiling model....')
model.compile(loss='categorical_crossentropy', optimizer=opt, metrics=['accuracy'])
return model
def Colorize(weights_path=None):
model = Sequential()
# input: 100x100 images with 3 channels -> (3, 100, 100) tensors.
# this applies 32 convolution filters of size 3x3 each.
model.add(Convolution2D(512, 1, 1, border_mode='valid',input_shape=(960,224,224)))
model.add(Activation('relu'))
model.add(normalization.BatchNormalization())
model.add(Convolution2D(256, 1, 1, border_mode='valid'))
model.add(Activation('relu'))
model.add(normalization.BatchNormalization())
model.add(Convolution2D(112, 1, 1, border_mode='valid'))
model.add(Activation('relu'))
model.add(normalization.BatchNormalization())
print "output shape: ",model.output_shape
#softmax
model.add(Reshape((112,224*224)))
print "output_shape after reshaped: ",model.output_shape
model.add(Activation('softmax'))
if weights_path:
model.load_weights(weights_path)
return model
def model(df, parent_id, go_id):
# Training
batch_size = 64
nb_epoch = 64
# Split pandas DataFrame
n = len(df)
split = 0.8
m = int(n * split)
train, test = df[:m], df[m:]
# train, test = train_test_split(
# labels, data, batch_size=batch_size)
train_label, train_data = train['labels'], train['data']
if len(train_data) < 100:
raise Exception("No training data for " + go_id)
test_label, test_data = test['labels'], test['data']
test_label_rep = test_label
train_data = train_data.as_matrix()
test_data = test_data.as_matrix()
train_data = numpy.hstack(train_data).reshape(train_data.shape[0], 8000)
test_data = numpy.hstack(test_data).reshape(test_data.shape[0], 8000)
shape = numpy.shape(train_data)
print('X_train shape: ', shape)
print('X_test shape: ', test_data.shape)
model = Sequential()
model.add(Dense(8000, activation='relu', input_dim=8000))
model.add(Highway())
model.add(Dense(1, activation='sigmoid'))
model.compile(
loss='binary_crossentropy', optimizer='rmsprop', class_mode='binary')
model_path = DATA_ROOT + parent_id + '/' + go_id + '.hdf5'
checkpointer = ModelCheckpoint(
filepath=model_path, verbose=1, save_best_only=True)
earlystopper = EarlyStopping(monitor='val_loss', patience=7, verbose=1)
model.fit(
X=train_data, y=train_label,
batch_size=batch_size, nb_epoch=nb_epoch,
show_accuracy=True, verbose=1,
validation_split=0.2,
callbacks=[checkpointer, earlystopper])
# Loading saved weights
print 'Loading weights'
model.load_weights(model_path)
pred_data = model.predict_classes(
test_data, batch_size=batch_size)
return classification_report(list(test_label_rep), pred_data)
def make_model_full(inshape, num_classes, weights_file=None):
model = Sequential()
model.add(KL.InputLayer(input_shape=inshape[1:]))
# model.add(KL.Conv2D(32, (3, 3), padding='same', input_shape=inshape[1:]))
model.add(KL.Conv2D(32, (3, 3), padding='same'))
model.add(KL.Activation('relu'))
model.add(KL.Conv2D(32, (3, 3)))
model.add(KL.Activation('relu'))
model.add(KL.MaxPooling2D(pool_size=(2, 2)))
model.add(KL.Dropout(0.25))
model.add(KL.Conv2D(64, (3, 3), padding='same'))
model.add(KL.Activation('relu'))
model.add(KL.Conv2D(64, (3, 3)))
model.add(KL.Activation('relu'))
model.add(KL.MaxPooling2D(pool_size=(2, 2)))
model.add(KL.Dropout(0.25))
model.add(KL.Flatten())
model.add(KL.Dense(512))
model.add(KL.Activation('relu'))
model.add(KL.Dropout(0.5))
model.add(KL.Dense(num_classes))
model.add(KL.Activation('softmax'))
if weights_file is not None and os.path.exists(weights_file):
model.load_weights(weights_file)
return model
def temporalNet(weights=None):
model = Sequential()
#3D convolutional layer with 32x32 optical flow as input
model.add(Convolution3D(30, 20, 17, 17, subsample=(4,2,2), input_shape=(1, 120,32,32)))
model.add(Activation(LeakyReLU()))
model.add(BatchNormalization())
model.add(MaxPooling3D(pool_size=(13, 2, 2), strides=(13,2, 2)))
model.add(Reshape((60, 4, 4)))
model.add(Convolution2D(100, 3, 3))
model.add(Activation(LeakyReLU()))
model.add(BatchNormalization())
model.add(MaxPooling2D((2, 2), strides=(2, 2)))
model.add(Flatten())
model.add(Dense(400))
model.add(Activation(LeakyReLU()))
model.add(Dropout(0.2))
model.add(Dense(50))
model.add(Activation(LeakyReLU()))
model.add(BatchNormalization())
model.add(Dense(4, activation='softmax'))
if weights:
model.load_weights(weights)
return model
def create_model(load_weights=False):
nn = Sequential()
nn.add(Convolution2D(32, 1, 3, 3, border_mode='same', activation='relu'))
nn.add(Convolution2D(32, 32, 3, 3, border_mode='same', activation='relu'))
nn.add(MaxPooling2D(poolsize=(2,2)))
nn.add(Dropout(0.25))
nn.add(Convolution2D(64, 32, 3, 3, border_mode='same', activation='relu'))
nn.add(Convolution2D(64, 64, 3, 3, border_mode='same', activation='relu'))
nn.add(MaxPooling2D(poolsize=(2,2)))
nn.add(Dropout(0.25))
nn.add(Flatten())
nn.add(Dense(64*7*7, 256, activation='relu'))
nn.add(Dropout(0.5))
nn.add(Dense(256,10, activation='softmax'))
sgd = SGD(lr=0.1, decay=1e-6, momentum=0.9, nesterov=True)
nn.compile(loss='categorical_crossentropy', optimizer=sgd)
if load_weights:
nn.load_weights('cnn_weights.hdf5')
return nn
def vgg_basic(img_size, weights_path = None, lr = 0.001):
'''
INPUT: img_size = size of images to train/ model was trained on
weights_path = path to get weights of trained model
OUTPUT: the fitted/unfitted model depending on if a weights path was
specified
A basic convolutional neural net. I found this one to have the best results.
'''
model = Sequential()
model.add(ZeroPadding2D((1,1),input_shape=(3, img_size, img_size)))
model.add(Convolution2D(64, 3, 3, activation='relu'))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(64, 3, 3, activation='relu'))
model.add(MaxPooling2D((2,2), strides=(2,2)))
model.add(Flatten())
model.add(Dense(1024, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(5, activation='softmax'))
if weights_path:
model.load_weights(weights_path)
adam = Adam(lr = lr)
model.compile(optimizer=adam,
loss='categorical_crossentropy')
return model
def logistic_regression(model_folder, layer, dimension, number_of_feature,
cost="binary_crossentropy", learning_rate=1e-6, dropout_rate=0.5, nepoch=10, activation="relu"):
model = Sequential()
model.add(Dense(dimension, input_dim=number_of_feature, init="uniform", activation=activation))
model.add(Dropout(dropout_rate))
for idx in range(0, layer-2, 1):
model.add(Dense(dimension, input_dim=dimension, init="uniform", activation=activation))
model.add(Dropout(dropout_rate))
model.add(Dense(1, init="uniform", activation="sigmoid"))
optimizer = RMSprop(lr=learning_rate, rho=0.9, epsilon=1e-06)
model.compile(loss=cost, optimizer=optimizer, metrics=['accuracy'])
filepath_model = get_newest_model(model_folder)
if filepath_model:
model.load_weights(filepath_model)
log("Load weights from {}".format(filepath_model), INFO)
else:
log("A new one model, {}".format(model_folder), INFO)
return model
def predict_ranking(evalFile, outFile):
X, qids, pids = load_data(evalFile)
input_dim = X[0].shape[1]
assert len(pids[0]) == len(X[0])
model = Sequential()
model.add(Dense(64, input_dim=input_dim, init='uniform', activation='relu'))
model.add(Dense(64, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(1, activation='sigmoid'))
model.compile(loss='binary_crossentropy', optimizer='rmsprop', metrics=['accuracy'])
weightsFile = '../model/weights.hdf5'
model.load_weights(weightsFile)
Y_p = []
for x in X:
Y_p.append(model.predict(x))
f = open(outFile, 'w')
for n, qid in enumerate(qids):
tupes = zip(Y_p[n], pids[n])
sortedTupes = sorted(tupes, key=lambda x: x[0], reverse=True)
for n, (y, pid) in enumerate(sortedTupes):
f.write('{}\tITER\t{}\t{}\t{}\tSOMEID\n'.format(qid, pid, n, 1001-n))
def M8(weights_path=None, input_shape=(1, 64, 64), n_output=None):
model = Sequential()
model.add(Convolution2D(
32, 3, 3, border_mode='same', input_shape=input_shape))
model.add(Activation('relu'))
model.add(Convolution2D(32, 3, 3))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.5))
model.add(Convolution2D(64, 3, 3, border_mode='same'))
model.add(Activation('relu'))
model.add(Convolution2D(64, 3, 3))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.5))
model.add(Convolution2D(128, 3, 3, border_mode='same'))
model.add(Activation('relu'))
model.add(Convolution2D(128, 3, 3))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.5))
model.add(Flatten())
model.add(Dense(1024))
model.add(Activation('relu'))
model.add(Dropout(0.5))
model.add(Dense(n_output))
model.add(Activation('softmax'))
if weights_path:
model.load_weights(weights_path)
return model
def get_nn_model(token_dict_size):
_logger.info('Initializing NN model with the following params:')
_logger.info('Input dimension: %s (token vector size)' % TOKEN_REPRESENTATION_SIZE)
_logger.info('Hidden dimension: %s' % HIDDEN_LAYER_DIMENSION)
_logger.info('Output dimension: %s (token dict size)' % token_dict_size)
_logger.info('Input seq length: %s ' % INPUT_SEQUENCE_LENGTH)
_logger.info('Output seq length: %s ' % ANSWER_MAX_TOKEN_LENGTH)
_logger.info('Batch size: %s' % SAMPLES_BATCH_SIZE)
model = Sequential()
seq2seq = SimpleSeq2seq(
input_dim=TOKEN_REPRESENTATION_SIZE,
input_length=INPUT_SEQUENCE_LENGTH,
hidden_dim=HIDDEN_LAYER_DIMENSION,
output_dim=token_dict_size,
output_length=ANSWER_MAX_TOKEN_LENGTH,
depth=1
)
model.add(seq2seq)
model.compile(loss='mse', optimizer='rmsprop')
model.save_weights(NN_MODEL_PATH)
# use previously saved model if it exists
_logger.info('Looking for a model %s' % NN_MODEL_PATH)
if os.path.isfile(NN_MODEL_PATH):
_logger.info('Loading previously calculated weights...')
model.load_weights(NN_MODEL_PATH)
_logger.info('Model is built')
return model
def C3D_Sports1M(weights_path=None):
model = Sequential()
# 1st layer group
model.add(Convolution3D(64, 3, 3, 3, activation='relu',
border_mode='same', name='conv1',
subsample=(1, 1, 1),
input_shape=(3, 16, 112, 112)))
model.add(MaxPooling3D(pool_size=(1, 2, 2), strides=(1, 2, 2),
border_mode='valid', name='pool1'))
# 2nd layer group
model.add(Convolution3D(128, 3, 3, 3, activation='relu',
border_mode='same', name='conv2',
subsample=(1, 1, 1)))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2),
border_mode='valid', name='pool2'))
# 3rd layer group
model.add(Convolution3D(256, 3, 3, 3, activation='relu',
border_mode='same', name='conv3a',
subsample=(1, 1, 1)))
model.add(Convolution3D(256, 3, 3, 3, activation='relu',
border_mode='same', name='conv3b',
subsample=(1, 1, 1)))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2),
border_mode='valid', name='pool3'))
# 4th layer group
model.add(Convolution3D(512, 3, 3, 3, activation='relu',
border_mode='same', name='conv4a',
subsample=(1, 1, 1)))
model.add(Convolution3D(512, 3, 3, 3, activation='relu',
border_mode='same', name='conv4b',
subsample=(1, 1, 1)))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2),
border_mode='valid', name='pool4'))
# 5th layer group
model.add(Convolution3D(512, 3, 3, 3, activation='relu',
border_mode='same', name='conv5a',
subsample=(1, 1, 1)))
model.add(Convolution3D(512, 3, 3, 3, activation='relu',
border_mode='same', name='conv5b',
subsample=(1, 1, 1)))
model.add(ZeroPadding3D(padding=(0, 1, 1)))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2),
border_mode='valid', name='pool5'))
model.add(Flatten())
# FC layers group
model.add(Dense(4096, activation='relu', name='fc6'))
model.add(Dropout(.5))
model.add(Dense(4096, activation='relu', name='fc7'))
model.add(Dropout(.5))
model.add(Dense(487, activation='softmax', name='fc8'))
if weights_path:
model.load_weights(weights_path)
return model
def vgg_train(weights=None):
print "Compiling VGG Model..."
model = Sequential()
# input: 64x64 images with 3 channels -> (3, 64, 64) tensors.
# this applies 32 convolution filters of size 3x3 each.
model.add(Convolution2D(32, 3, 3, border_mode='valid', input_shape=(3,64,64)))
model.add(Activation('relu'))
model.add(Convolution2D(32, 3, 3))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Convolution2D(64, 3, 3, border_mode='valid'))
model.add(Activation('relu'))
model.add(Convolution2D(64, 3, 3))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.50))
model.add(Flatten())
# Note: Keras does automatic shape inference.
model.add(Dense(256,W_regularizer=WeightRegularizer(l1=1e-6,l2=1e-6)))
model.add(Activation('relu'))
model.add(Dropout(0.50))
model.add(Dense(200,W_regularizer=WeightRegularizer(l1=1e-5,l2=1e-5)))
model.add(Activation('softmax'))
if weights!=None:
model.load_weights(weights)
return model
class Model: def __init__(self, init='normal', activation='relu', batch_size=32, lr=1e-2, load=None): self.batch_size = batch_size self.model = Sequential() self.model.add(Convolution1D(8, 1, input_shape=(2, 4), init=init)) self.model.add(Activation(activation)) self.model.add(Flatten()) self.model.add(Dense(8, init=init)) self.model.add(BatchNormalization()) self.model.add(Activation(activation)) self.model.add(Dense(8, init=init)) self.model.add(BatchNormalization()) self.model.add(Activation(activation)) self.model.add(Dense(2, init=init)) if load != None: self.model.load_weights(load) self.model.compile(SGD(lr=lr), loss='mse') def predict(self, X): return self.model.predict(X.reshape((1,) + X.shape))[0] def learn(self, X, y): return self.model.fit(X, y, nb_epoch=1, batch_size=self.batch_size, shuffle=True, verbose=0) def save(self, filename): return self.model.save_weights(filename, overwrite=True)
def part4(weights):
global X_train, Y_train, X_test, Y_test
size = 3
model = Sequential()
model.add(Convolution2D(32, size, size, border_mode="same", input_shape=(img_channels, img_rows, img_cols)))
model.add(Activation("relu"))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.5))
model = add_top_layer(model)
model.load_weights(weights)
model = copy_freeze_model(model, 4)
# add top dense layer
model.add(Flatten())
model.add(Dense(512))
model.add(Activation("relu"))
model.add(Dropout(0.5))
model.add(Dense(10))
model.add(Activation("softmax"))
X_train, Y_train, X_test, Y_test = load_data(10)
model = train(model, auto=False)
print("Classification rate %02.5f" % (model.evaluate(X_test, Y_test, show_accuracy=True)[1]))
def temporalNet(weights=None):
model = Sequential()
model.add(Convolution3D(30, 20, 17, 17, activation='relu', trainable=False, subsample=(4,2,2), input_shape=(1, 120,32,32)))
model.add(MaxPooling3D(pool_size=(13, 2, 2), strides=(13,2, 2), trainable=False))
model.add(Reshape((60, 4, 4)))
model.add(Convolution2D(100, 3, 3, activation='relu', trainable=False))
model.add(MaxPooling2D((2, 2), strides=(2, 2), trainable=False))
model.add(Flatten())
model.add(Dense(400, activation='relu', trainable=False))
model.add(Dense(50, activation='relu'))
#model.add(Dense(14, activation='relu'))
model.add(Dense(1, activation='relu'))
if weights:
model.load_weights(weights)
return model
def get_conv(input_shape=(48, 48, 1), filename=None):
model = Sequential()
# model.add(Lambda(lambda x: (x-np.mean(x))/np.std(x),input_shape=input_shape, output_shape=input_shape))
model.add(Conv2D(32, (3, 3), activation='relu', name='conv1',
input_shape=input_shape, padding="same"))
model.add(Conv2D(32, (3, 3), activation='relu',
name='conv2', padding="same"))
model.add(MaxPooling2D(pool_size=(4, 4)))
model.add(Dropout(0.25))
model.add(Conv2D(64, (3, 3), name='conv3', padding='same'))
model.add(Activation('relu'))
model.add(Conv2D(64, (3, 3), name='conv4', padding='same'))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(4, 4)))
model.add(Dropout(0.25))
model.add(Conv2D(256, (3, 3), activation="relu",
name="dense1")) # This was Dense(128)
model.add(Dropout(0.5))
# This was Dense(1)
model.add(Conv2D(1, (1, 1), name="dense2", activation="tanh"))
if filename:
model.load_weights(filename)
return model
def spatialNet(weights=None):
model = Sequential()
leakyrelu = ELU()
model.add(Convolution3D(30, 10, 17, 17, subsample=(2,2,2), trainable=False, input_shape=(1, 60,32,32)))
model.add(Activation(leakyrelu))
model.add(BatchNormalization(mode=2))
model.add(MaxPooling3D(pool_size=(13, 2, 2),trainable=False, strides=(13,2, 2)))
model.add(Reshape((60, 4, 4)))
model.add(Convolution2D(100, 3, 3, trainable=False))
model.add(Activation(leakyrelu))
model.add(BatchNormalization(mode=2))
model.add(MaxPooling2D((2, 2), strides=(2, 2), trainable=False))
model.add(Flatten())
model.add(Dense(400, trainable=False))
model.add(Activation(leakyrelu))
model.add(BatchNormalization(mode=2))
model.add(Dense(50))
model.add(Activation(leakyrelu))
model.add(BatchNormalization(mode=2))
model.add(Dense(1, activation='sigmoid'))
model.add(BatchNormalization(mode=2))
if weights:
model.load_weights(weights)
return model
def __init__(self, restore=None, session=None, Dropout=Dropout, num_labels=10):
self.num_channels = 1
self.image_size = 28
self.num_labels = num_labels
model = Sequential()
nb_filters = 64
layers = [Conv2D(nb_filters, (5, 5), strides=(2, 2), padding="same",
input_shape=(28, 28, 1)),
Activation('relu'),
Conv2D(nb_filters, (3, 3), strides=(2, 2), padding="valid"),
Activation('relu'),
Conv2D(nb_filters, (3, 3), strides=(1, 1), padding="valid"),
Activation('relu'),
Flatten(),
Dense(32),
Activation('relu'),
Dropout(.5),
Dense(num_labels)]
for layer in layers:
model.add(layer)
if restore != None:
model.load_weights(restore)
self.model = model
def home_made_convnet(Xshape, Yshape):
print 'initializing model...'
model = Sequential()
model.add(ZeroPadding2D((2, 2), input_shape=(Xshape))) # 0
model.add(Convolution2D(64, 5, 5, activation='relu')) # 1
model.add(Convolution2D(64, 5, 5, activation='relu')) # 3
model.add(MaxPooling2D((2, 2), strides=(2, 2))) # 4
model.add(Convolution2D(128, 3, 3, activation='relu')) # 6
model.add(Convolution2D(128, 3, 3, activation='relu')) # 8
model.add(MaxPooling2D((2, 2), strides=(2, 2))) # 9
model.add(Flatten()) # 31
model.add(Dense(256, activation='relu')) # 32
model.add(Dropout(0.5)) # 33
model.add(Dense(256, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(Yshape, activation='softmax'))
model.load_weights('weights_f.h5')
sgd = SGD(lr=0.001, decay=1e-6, momentum=0.9, nesterov=True)
model.compile(optimizer=sgd,
loss='categorical_crossentropy',
metrics=['accuracy'])
return model
def test_sequential_model_saving():
model = Sequential()
model.add(Dense(2, input_dim=3))
model.add(Dense(3))
model.compile(loss='mse', optimizer='rmsprop', metrics=['acc'])
x = np.random.random((1, 3))
y = np.random.random((1, 3))
model.train_on_batch(x, y)
out = model.predict(x)
fname = 'tmp_' + str(np.random.randint(10000)) + '.h5'
save_model(model, fname)
new_model = load_model(fname)
out2 = new_model.predict(x)
assert_allclose(out, out2, atol=1e-05)
# test that new updates are the same with both models
x = np.random.random((1, 3))
y = np.random.random((1, 3))
model.train_on_batch(x, y)
new_model.train_on_batch(x, y)
out = model.predict(x)
out2 = new_model.predict(x)
assert_allclose(out, out2, atol=1e-05)
# test load_weights on model file
model.load_weights(fname)
os.remove(fname)
def build(width, height, depth, num_classes, weights_path=None, dropout=True):
model = Sequential()
# First set of CONV => RELU => POOL.
model.add(Convolution2D(20, 5, 5, border_mode='same',
input_shape=(height, width, depth)))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2)))
# Second set of CONV => RELU => POOL.
model.add(Convolution2D(50, 5, 5, border_mode='same'))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2)))
# Last set of FC => RELU layers
model.add(Flatten())
model.add(Dense(500))
model.add(Activation('relu'))
if dropout:
model.add(Dropout(.5))
# softmax classifier
model.add(Dense(num_classes))
model.add(Activation('softmax'))
# If a weights path is supplied (inicating that the model was
# pre-trained), then load the weights
if weights_path is not None:
model.load_weights(weights_path)
# return the constructed network architecture
return model
def CIFAR_10(img_rows, img_cols, img_channels=3, nb_classes=5, weights_path=None):
model = Sequential()
model.add(Convolution2D(32, 3, 3, border_mode='same',
input_shape=(img_channels, img_rows, img_cols)))
model.add(Activation('relu'))
model.add(Convolution2D(32, 3, 3))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Convolution2D(64, 3, 3, border_mode='same'))
model.add(Activation('relu'))
model.add(Convolution2D(64, 3, 3))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Flatten())
model.add(Dense(512))
model.add(Activation('relu'))
model.add(Dropout(0.5))
model.add(Dense(nb_classes))
model.add(Activation('softmax'))
if weights_path:
model.load_weights(weights_path)
return model
def Model(weights_path=None):
model = Sequential()
model.add(GaussianNoise(0.01, input_shape=(1, img_rows, img_cols)))
model.add(Convolution2D(nb_filters[0], nb_conv, nb_conv,
border_mode='valid'))
model.add(Activation('relu'))
model.add(Convolution2D(nb_filters[1], nb_conv, nb_conv))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(nb_pool, nb_pool)))
model.add(Dropout(0.25))
model.add(Convolution2D(nb_filters[2], nb_conv, nb_conv))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(nb_pool, nb_pool)))
model.add(Dropout(0.25))
model.add(Convolution2D(nb_filters[3], nb_conv, nb_conv))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(nb_pool, nb_pool)))
model.add(Dropout(0.25))
model.add(Flatten())
model.add(Dense(128))
model.add(Activation('relu'))
model.add(Dropout(0.5))
model.add(Dense(nb_classes))
model.add(Activation('softmax'))
if weights_path:
model.load_weights(weights_path)
return model
