def preprocess():
# # Splitting dataset in train,test and validate
# x_train,y_train,x_test,y_test,valid_set_x,valid_set_y = split(training_per=0.8,test_per=0.1,validation_per=0.1)
# x_train = x_train.reshape(x_train.shape[0],32,32,1)
# x_test = x_test.reshape(x_test.shape[0],32,32,1)
# valid_set_x = valid_set_x.reshape(valid_set_x.shape[0],32,32,1)
# # print(x_train.shape)
# # print(x_test.shape)
# # print(valid_set_x.shape)
# hot_enc = one_hot_encoding()
# y_train = [hot_enc[int(x)] for x in y_train]
# y_train = np.asarray(y_train)
# y_train = y_train.reshape(x_train.shape[0], 10)
# y_test = [hot_enc[int(x)] for x in y_test]
# y_test = np.asarray(y_test)
# y_test = y_test.reshape(x_test.shape[0], 10)
# y_val = [hot_enc[int(x)] for x in valid_set_y]
# y_val = np.asarray(y_val)
# valid_set_y = y_val.reshape(valid_set_x.shape[0], 10)
import tflearn.datasets.oxflower17 as oxflower17
X, Y = oxflower17.load_data(one_hot=True)
x_train, x_test_pre, y_train, y_test_pre = train_test_split(
X, Y, test_size=0.20, random_state=42)
x_test, x_validation, y_test, y_validation = train_test_split(
x_test_pre, y_test_pre, test_size=0.1)
return x_train, y_train, x_test, y_test, x_validation, y_validation
def alexnet():
X, Y = oxflower17.load_data(one_hot=True, resize_pics=(227, 227))
# Building 'AlexNet'
network = input_data(shape=[None, 227, 227, 3])
network = conv_2d(network, 96, 11, strides=4, activation='relu')
network = max_pool_2d(network, 3, strides=2)
network = local_response_normalization(network)
network = conv_2d(network, 256, 5, activation='relu')
network = max_pool_2d(network, 3, strides=2)
network = local_response_normalization(network)
network = conv_2d(network, 384, 3, activation='relu')
network = conv_2d(network, 384, 3, activation='relu')
network = conv_2d(network, 256, 3, activation='relu')
network = max_pool_2d(network, 3, strides=2)
network = local_response_normalization(network)
network = fully_connected(network, 4096, activation='tanh')
network = dropout(network, 0.5)
network = fully_connected(network, 4096, activation='tanh')
network = dropout(network, 0.5)
network = fully_connected(network, 17, activation='softmax')
network = regression(network, optimizer='momentum',
loss='categorical_crossentropy',
learning_rate=0.001)
# Training
model = tflearn.DNN(network, checkpoint_path='model_alexnet',
max_checkpoints=1, tensorboard_verbose=2)
model.fit(X, Y, n_epoch=1000, validation_set=0.1, shuffle=True,
show_metric=True, batch_size=64, snapshot_step=200,
snapshot_epoch=False, run_id='alexnet')
def pre_train():
if os.path.isfile('models/pre_train.model.index'):
print("Previous trained model exist.")
return
X, Y = oxflower17.load_data(one_hot=True, resize_pics=(227, 227))
net = create_alexnet(17)
model = tflearn.DNN(net,
checkpoint_path='ckps/pre_train.ckp',
max_checkpoints=1,
tensorboard_verbose=2,
tensorboard_dir='tmp/pre_train_logs/')
if os.path.isfile('models/pre_train.model'):
model.load('models/pre_train.model')
model.fit(X,
Y,
n_epoch=100,
validation_set=0.1,
shuffle=True,
show_metric=True,
batch_size=64,
snapshot_step=200,
snapshot_epoch=False,
run_id='pre_train')
# Save the model
model.save('models/pre_train.model')
def unmatch_data(X_data,
y_data,
unmatch_per,
oriset='cifar10',
name='oxflower17'):
size1 = round(train_data_num * (1 - unmatch_per))
X1, y1 = random_data(X_data, y_data, size1)
size2 = round(train_data_num * unmatch_per)
if name == 'FashionMNIST':
(X_, y_), (X_test1, y_test1) = fashion_mnist.load_data()
X2, y2 = random_data(X_, y_, size2)
if name == 'oxflower17':
X_, y_ = oxflower17.load_data(dirname="17flowers", one_hot=False)
X_ = np.array([cv2.resize(i, (32, 32)) for i in X_])
y_ = np.array([random.randint(0, 9) for _ in range(y_.shape[0])])
print("oxflower y shape: ", y_.shape)
if size2 > y_.shape[0]:
rnd = size2 // y_.shape[0] + 1
X_new = X_
y_new = y_
for _ in range(rnd):
X_new = np.concatenate([X_new, X_])
y_new = np.concatenate([y_new, y_])
X2, y2 = random_data(X_new, y_new, size2)
print("Unmatched dataset shape: ", X2.shape)
return np.concatenate([X1, X2]), np.concatenate([y1, y2])
def vggnet():
X, Y = oxflower17.load_data(one_hot=True, resize_pics=(227, 227))
input = input_data(shape=[None, 227, 227, 3], name='input')
conv1 = conv_2d(input, 64, 3, activation='relu')
conv2 = conv_2d(conv1, 64, 3, activation='relu')
pool1 = max_pool_2d(conv2, 2, strides=2)
conv3 = conv_2d(pool1, 128, 3, activation='relu')
conv4 = conv_2d(conv3, 128, 3, activation='relu')
pool2 = max_pool_2d(conv4, 2, strides=2)
conv5 = conv_2d(pool2, 256, 3, activation='relu')
conv6 = conv_2d(conv5, 256, 3, activation='relu')
conv7 = conv_2d(conv6, 256, 3, activation='relu')
pool3 = max_pool_2d(conv7, 2, strides=2)
conv8 = conv_2d(pool3, 512, 3, activation='relu')
conv9 = conv_2d(conv8, 512, 3, activation='relu')
conv10 = conv_2d(conv9, 512, 3, activation='relu')
pool4 = max_pool_2d(conv10, 2, strides=2)
conv11 = conv_2d(pool4, 512, 3, activation='relu')
conv12 = conv_2d(conv11, 512, 3, activation='relu')
conv13 = conv_2d(conv12, 512, 3, activation='relu')
pool5 = max_pool_2d(conv13, 2, strides=2)
lrn = local_response_normalization(pool5)
fc1 = fully_connected(lrn, 4096, activation='tanh')
dr1 = dropout(fc1, 0.5)
fc2 = fully_connected(dr1, 4096, activation='relu')
dr2 = dropout(fc2, 0.5)
fc3 = fully_connected(dr2, 17, activation='softmax')
# 声明优化算法、损失函数、学习率等
network = regression(fc3,
optimizer='rmsprop',
loss='categorical_crossentropy',
learning_rate=0.01)
# Training
model = tflearn.DNN(network,
checkpoint_path='model_vgg',
max_checkpoints=10,
tensorboard_verbose=2,
tensorboard_dir='logs/vgg')
# n_epoch=10表示整个训练数据集将会用10遍，
# batch_size=16表示一次用16个数据计算参数的更新
model.fit(X,
Y,
n_epoch=1,
validation_set=0.1,
shuffle=True,
show_metric=True,
batch_size=64,
snapshot_step=200,
snapshot_epoch=False,
run_id='vgg')
def get_oxflower17_data(num_train=1000, num_validation=180, num_test=180):
# Load the raw oxflower17 data
X, y = oxflower17.load_data()
# Shuffle the data
# Subsample the data
return X_train, y_train, X_val, y_val, X_test, y_test
def vggnet():
X, Y = oxflower17.load_data(one_hot=True, resize_pics=(227, 227))
# Building 'VGG Network'
network = input_data(shape=[None, 227, 227, 3])
network = conv_2d(network, 64, 3, activation='relu')
network = conv_2d(network, 64, 3, activation='relu')
network = max_pool_2d(network, 2, strides=2)
network = conv_2d(network, 128, 3, activation='relu')
network = conv_2d(network, 128, 3, activation='relu')
network = max_pool_2d(network, 2, strides=2)
network = conv_2d(network, 256, 3, activation='relu')
network = conv_2d(network, 256, 3, activation='relu')
network = conv_2d(network, 256, 3, activation='relu')
network = max_pool_2d(network, 2, strides=2)
network = conv_2d(network, 512, 3, activation='relu')
network = conv_2d(network, 512, 3, activation='relu')
network = conv_2d(network, 512, 3, activation='relu')
network = max_pool_2d(network, 2, strides=2)
network = conv_2d(network, 512, 3, activation='relu')
network = conv_2d(network, 512, 3, activation='relu')
network = conv_2d(network, 512, 3, activation='relu')
network = max_pool_2d(network, 2, strides=2)
network = fully_connected(network, 4096, activation='relu')
network = dropout(network, 0.5)
network = fully_connected(network, 4096, activation='relu')
network = dropout(network, 0.5)
network = fully_connected(network, 17, activation='softmax')
network = regression(network,
optimizer='rmsprop',
loss='categorical_crossentropy',
learning_rate=0.0001)
# Training
model = tflearn.DNN(network,
checkpoint_path='model_vgg',
max_checkpoints=1,
tensorboard_verbose=0)
model.fit(X,
Y,
n_epoch=500,
shuffle=True,
show_metric=True,
batch_size=32,
snapshot_step=500,
snapshot_epoch=False,
run_id='vgg')
def get_data_oxford_flowers(): import tflearn.datasets.oxflower17 as oxflower17 X, Y = oxflower17.load_data(one_hot = True, resize_pics = (227, 227)) split_percentage = 80 split_index = int(X.shape[0]/(100/split_percentage)) x_train = np.array(X[:split_index]) x_val = np.array(X[split_index:]) y_train = np.array(Y[:split_index]) y_val = np.array(Y[split_index:]) return x_train, x_val, y_train, y_val
def train():
x, y = oxflower17.load_data(one_hot=True, resize_pics=(227, 227))
model = tflearn_model()
model.fit(x,
y,
n_epoch=1000,
validation_set=0.1,
shuffle=True,
show_metric=True,
batch_size=512,
snapshot_step=200,
snapshot_epoch=False,
run_id='alexnet_oxflowers17')
def load_data():
Xtrain, Ytrain = oxflower17.load_data(one_hot=True)
input_shape = (Xtrain.shape[1], Xtrain.shape[2], Xtrain.shape[3]
) # (224,224,3)
num_classes = Ytrain.shape[1] # 17
print("Training input %s" % str(Xtrain.shape))
print("Training output %s" % str(Ytrain.shape))
#print("Test input %s" %str(Xtest.shape))
#print("Test output %s" %str(Ytest.shape))
print("Input shape: %s" % str(input_shape))
print("Number of classes: %d" % num_classes)
return [Xtrain, Ytrain, input_shape, num_classes]
def load_and_transform_data(test_size=0.20):
"""
Returns oxflower17 dataset by splitting them to train and test sets.
"""
X, Y = oxflower17.load_data(one_hot=True)
print("X's shape: ", X.shape)
print("Y's shape: ", Y.shape)
X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=test_size)
print("X_train's shape: ", X_train.shape)
print("Y_train's shape: ", Y_train.shape)
return X_train, X_test, Y_train, Y_test
def vggnet():
"""
Returns:None
"""
X, Y = oxflower17.load_data(resize_pics=(227, 227), one_hot=True)
# Building 'VGG Network'
inputs = Input(shape=(227, 227, 3))
network = Conv2D(64, (3, 3), activation='relu')(inputs)
network = Conv2D(64, (3, 3), activation='relu')(network)
network = MaxPool2D(pool_size=(2, 2), strides=2)(network)
network = Conv2D(128, (3, 3), activation='relu')(network)
network = Conv2D(128, (3, 3), activation='relu')(network)
network = MaxPool2D(pool_size=(2, 2), strides=2)(network)
network = Conv2D(256, (3, 3), activation='relu')(network)
network = Conv2D(256, (3, 3), activation='relu')(network)
network = Conv2D(256, (3, 3), activation='relu')(network)
network = MaxPool2D(pool_size=(2, 2), strides=2)(network)
network = Conv2D(512, (3, 3), activation='relu')(network)
network = Conv2D(512, (3, 3), activation='relu')(network)
network = Conv2D(512, (3, 3), activation='relu')(network)
network = MaxPool2D(pool_size=(2, 2), strides=2)(network)
network = Conv2D(512, (3, 3), activation='relu')(network)
network = Conv2D(512, (3, 3), activation='relu')(network)
network = Conv2D(512, (3, 3), activation='relu')(network)
network = MaxPool2D(pool_size=(2, 2), strides=2)(network)
network = Flatten()(network)
network = Dense(4096, activation='relu')(network)
network = Dropout(0.5)(network)
network = Dense(4096, activation='relu')(network)
network = Dropout(0.5)(network)
predictions = Dense(17, activation='softmax')(network)
# Training
model = Model(inputs=inputs, outputs=predictions)
model.compile(optimizer='rmsprop',
loss='categorical_crossentropy',
metrics=['accuracy'])
model.fit(X, Y, n_epoch=500, batch_size=32, validation_split=0.2)
def main():
x, y = oxflower17.load_data(one_hot=True)
print('input X shape: ', x.shape)
print('label shape ', y.shape)
X_train, X_test, Y_train, Y_test = train_test_split(x,
y,
test_size=0.2,
random_state=42)
train_accuracy, test_accuracy, parameters = model(X_train,
Y_train,
X_test,
Y_test,
num_epochs=10)
def main():
x, y = oxflower17.load_data(one_hot=True)
AlexNet = create_model(img_shape=(224, 224, 3))
print(AlexNet.summary())
AlexNet.compile(loss='categorical_crossentropy',
optimizer='adam',
metrics=['accuracy'])
AlexNet.fit(x,
y,
batch_size=64,
epochs=100,
verbose=1,
validation_split=0.1,
shuffle=True)
def get_data(self):
'''
Each image is represented as a 224x224 pixel RGB colored image. Therefore,
each image is [3][224][224]float matrix. The dataset is already
pre-processed so all values are (0, 1).
Each label is a binary [17]float where each image is assigned to one of the 17
classes of flowers.
'''
self.data_x, self.data_y = oxflower17.load_data(dirname="data/17flowers", one_hot=True)
self.image_width, self.image_height, self.color_depth = \
self.data_x.shape[1], self.data_x.shape[2], self.data_x.shape[3]
self.n_classes = self.data_y.shape[1]
print('[+] Loaded dataset: ')
print('[+] Data point dimensions: ', self.data_x.shape)
print('[+] Label dimensions: ', self.data_y.shape)
def vggnet():
X, Y = oxflower17.load_data(one_hot=True,resize_pics=(227, 227))
# Building 'VGG Network'
network = input_data(shape=[None, 227, 227, 3])
network = conv_2d(network, 64, 3, activation='relu')
network = conv_2d(network, 64, 3, activation='relu')
network = max_pool_2d(network, 2, strides=2)
network = conv_2d(network, 128, 3, activation='relu')
network = conv_2d(network, 128, 3, activation='relu')
network = max_pool_2d(network, 2, strides=2)
network = conv_2d(network, 256, 3, activation='relu')
network = conv_2d(network, 256, 3, activation='relu')
network = conv_2d(network, 256, 3, activation='relu')
network = max_pool_2d(network, 2, strides=2)
network = conv_2d(network, 512, 3, activation='relu')
network = conv_2d(network, 512, 3, activation='relu')
network = conv_2d(network, 512, 3, activation='relu')
network = max_pool_2d(network, 2, strides=2)
network = conv_2d(network, 512, 3, activation='relu')
network = conv_2d(network, 512, 3, activation='relu')
network = conv_2d(network, 512, 3, activation='relu')
network = max_pool_2d(network, 2, strides=2)
network = fully_connected(network, 4096, activation='relu')
network = dropout(network, 0.5)
network = fully_connected(network, 4096, activation='relu')
network = dropout(network, 0.5)
network = fully_connected(network, 17, activation='softmax')
network = regression(network, optimizer='rmsprop',
loss='categorical_crossentropy',
learning_rate=0.0001)
# Training
model = tflearn.DNN(network, checkpoint_path='model_vgg',
max_checkpoints=1, tensorboard_verbose=0)
model.fit(X, Y, n_epoch=500, shuffle=True,
show_metric=True, batch_size=32, snapshot_step=500,
snapshot_epoch=False, run_id='vgg')
def get_oxflower17_data(num_training=1000, num_validation=180, num_test=180):
# Load the raw oxflower17 data
X, y = oxflower17.load_data()
# Subsample the data
mask = range(num_training)
X_train = X[mask]
y_train = y[mask]
mask = range(num_training, num_training+num_validation)
X_val = X[mask]
y_val = y[mask]
mask = range(num_training+num_validation, num_training+num_validation+num_test)
X_test = X[mask]
y_test = y[mask]
# Normalize the data: subtract the mean image
mean_image = np.mean(X_train, axis=0)
X_train -= mean_image
X_val -= mean_image
X_test -= mean_image
return X_train, y_train, X_val, y_val, X_test, y_test
def alexnet():
"""
Returns:None
"""
X, Y = oxflower17.load_data(one_hot=True, resize_pics=(227, 227))
# Building 'AlexNet'
inputs = Input(shape=(227, 227, 3))
network = Conv2D(96, (11, 11), strides=4, activation='relu')(inputs)
network = MaxPool2D(pool_size=(2, 2))(network)
network = BatchNormalization()(network)
network = Dropout(0.25)(network)
network = Conv2D(256, (5, 5), activation='relu')(network)
network = MaxPool2D(pool_size=(3, 3))(network)
network = BatchNormalization()(network)
network = Dropout(0.25)(network)
network = Conv2D(384, (3, 3), activation='relu')(network)
network = Conv2D(384, (3, 3), activation='relu')(network)
network = Conv2D(256, (3, 3), activation='relu')(network)
network = MaxPool2D(pool_size=(3, 3), strides=2)(network)
network = BatchNormalization()(network)
network = Dropout(0.25)(network)
network = Flatten()(network)
network = Dense(4096, activation='tanh')(network)
network = Dropout(0.5)(network)
network = Dense(4096, activation='tanh')(network)
network = Dropout(0.5)(network)
predictions = Dense(17, activation='softmax')(network)
# Training
model = Model(inputs=inputs, outputs=predictions)
model.compile(optimizer='momentum',
loss='categorical_crossentropy',
metrics=['accuracy'])
model.fit(X, Y, epochs=1000, validation_set=0.1, batch_size=64)
tower_conv1_2 = relu(batch_normalization(conv_2d(tower_conv1_1, 256, [3, 1], bias=False, name='Conv2d_0c_3x1')))
tower_mixed = merge([tower_conv, tower_conv1_2], mode='concat', axis=3)
tower_out = relu(batch_normalization(
conv_2d(tower_mixed, net.get_shape()[3], 1, bias=False, activation=None, name='Conv2d_1x1')))
net += scale * tower_out
if activation:
if isinstance(activation, str):
net = activations.get(activation)(net)
elif hasattr(activation, '__call__'):
net = activation(net)
else:
raise ValueError("Invalid Activation.")
return net
X, Y = oxflower17.load_data(one_hot=True, resize_pics=(299, 299))
num_classes = 17
dropout_keep_prob = 0.8
network = input_data(shape=[None, 299, 299, 3])
conv1a_3_3 = relu(batch_normalization(
conv_2d(network, 32, 3, strides=2, bias=False, padding='VALID', activation=None, name='Conv2d_1a_3x3')))
conv2a_3_3 = relu(
batch_normalization(conv_2d(conv1a_3_3, 32, 3, bias=False, padding='VALID', activation=None, name='Conv2d_2a_3x3')))
conv2b_3_3 = relu(batch_normalization(conv_2d(conv2a_3_3, 64, 3, bias=False, activation=None, name='Conv2d_2b_3x3')))
maxpool3a_3_3 = max_pool_2d(conv2b_3_3, 3, strides=2, padding='VALID', name='MaxPool_3a_3x3')
conv3b_1_1 = relu(batch_normalization(
conv_2d(maxpool3a_3_3, 80, 1, bias=False, padding='VALID', activation=None, name='Conv2d_3b_1x1')))
conv4a_3_3 = relu(batch_normalization(
conv_2d(conv3b_1_1, 192, 3, bias=False, padding='VALID', activation=None, name='Conv2d_4a_3x3')))
# Job id run_id = 'alexnet_oxflowers17_' + optimizer_name + str(learning_rate) # Device gpu = '/gpu:0' cpu = '/cpu:0' # Checkpoint & snapshot check_path = 'model_' + optimizer_name + str(learning_rate) max_checkpoints = 10 snapshot_step=200 is_snapshot_epoch=False # Dataset import tflearn.datasets.oxflower17 as oxflower17 X, Y = oxflower17.load_data(dirname='/home/jiawei/dataset/17flowers/', one_hot=True, resize_pics=(227, 227)) tflearn.config.init_graph(log_device=True, soft_placement=True) # Building 'AlexNet' network = input_data(shape=[None, 227, 227, 3]) network = conv_2d(network, 96, 11, strides=4, activation='relu') network = max_pool_2d(network, 3, strides=2) network = local_response_normalization(network) network = conv_2d(network, 256, 5, activation='relu') network = max_pool_2d(network, 3, strides=2) network = local_response_normalization(network) network = conv_2d(network, 384, 3, activation='relu') network = conv_2d(network, 384, 3, activation='relu') network = conv_2d(network, 256, 3, activation='relu') network = max_pool_2d(network, 3, strides=2)
# Applying VGG 16-layers convolutional network to Oxford's 17 Category Flower
# Dataset classification task.
import tflearn
import os
import tflearn.datasets.oxflower17 as oxflower17
X, Y = oxflower17.load_data(one_hot=True)
def run():
net = tflearn.input_data(shape=[None, 224, 224, 3])
net = tflearn.conv_2d(net, 64, 3, activation='relu')
net = tflearn.conv_2d(net, 64, 3, activation='relu')
net = tflearn.max_pool_2d(net, 2)
net = tflearn.conv_2d(net, 128, 3, activation='relu')
net = tflearn.conv_2d(net, 128, 3, activation='relu')
net = tflearn.max_pool_2d(net, 2)
net = tflearn.conv_2d(net, 256, 3, activation='relu')
net = tflearn.conv_2d(net, 256, 3, activation='relu')
net = tflearn.conv_2d(net, 256, 3, activation='relu')
net = tflearn.max_pool_2d(net, 2)
net = tflearn.conv_2d(net, 512, 3, activation='relu')
net = tflearn.conv_2d(net, 512, 3, activation='relu')
net = tflearn.conv_2d(net, 512, 3, activation='relu')
net = tflearn.max_pool_2d(net, 2)
net = tflearn.conv_2d(net, 512, 3, activation='relu')
net = tflearn.conv_2d(net, 512, 3, activation='relu')
return shuffled_dataset, shuffled_labels
# ######################################################
# 定义模型参数
# ######################################################
tf.flags.DEFINE_string('path', r'../17flowers', '经典数据集地址')
tf.flags.DEFINE_float('learning_rate', 0.002, '学习率')
tf.flags.DEFINE_float('dropout', 1, '每层输出DROPOUT的大小')
tf.flags.DEFINE_integer('batch_size', 32, '小批量梯度下降的批量大小')
tf.flags.DEFINE_float('sample', 0.1, '取样的数目')
tf.flags.DEFINE_integer('num_epoch', 1000, '训练几轮')
tf.flags.DEFINE_integer('num_class', 17, '一共多少类')
FLAGS = tf.flags.FLAGS
train_data, train_label = oxflower17.load_data(dirname=FLAGS.path,
one_hot=True)
train_data, train_label = mess_dataset_order(train_data,
train_label,
dimention=train_data.shape[1])
sample = int(FLAGS.sample * train_data.shape[0])
x, y = train_data[:sample, :, :, :], train_label[:sample]
inputs = tf.placeholder(dtype=tf.float32,
shape=[None, 224, 224, 3],
name='input')
output = tf.placeholder(tf.float32, [None, 17], name='output')
out, _ = inception_v3(inputs=inputs,
num_classes=FLAGS.num_class,
is_training=True,
tower_conv1_0 = relu(batch_normalization(conv_2d(net, 192, 1, bias=False, activation=None, name='Conv2d_0a_1x1')))
tower_conv1_1 = relu(batch_normalization(conv_2d(tower_conv1_0, 224, [1,3], bias=False, activation=None, name='Conv2d_0b_1x3')))
tower_conv1_2 = relu(batch_normalization(conv_2d(tower_conv1_1, 256, [3,1], bias=False, name='Conv2d_0c_3x1')))
tower_mixed = merge([tower_conv,tower_conv1_2], mode='concat', axis=3)
tower_out = relu(batch_normalization(conv_2d(tower_mixed, net.get_shape()[3], 1, bias=False, activation=None, name='Conv2d_1x1')))
net += scale * tower_out
if activation:
if isinstance(activation, str):
net = activations.get(activation)(net)
elif hasattr(activation, '__call__'):
net = activation(net)
else:
raise ValueError("Invalid Activation.")
return net
X, Y = oxflower17.load_data(one_hot=True, resize_pics=(299, 299))
num_classes = 17
dropout_keep_prob = 0.8
network = input_data(shape=[None, 299, 299, 3])
conv1a_3_3 = relu(batch_normalization(conv_2d(network, 32, 3, strides=2, bias=False, padding='VALID',activation=None,name='Conv2d_1a_3x3')))
conv2a_3_3 = relu(batch_normalization(conv_2d(conv1a_3_3, 32, 3, bias=False, padding='VALID',activation=None, name='Conv2d_2a_3x3')))
conv2b_3_3 = relu(batch_normalization(conv_2d(conv2a_3_3, 64, 3, bias=False, activation=None, name='Conv2d_2b_3x3')))
maxpool3a_3_3 = max_pool_2d(conv2b_3_3, 3, strides=2, padding='VALID', name='MaxPool_3a_3x3')
conv3b_1_1 = relu(batch_normalization(conv_2d(maxpool3a_3_3, 80, 1, bias=False, padding='VALID',activation=None, name='Conv2d_3b_1x1')))
conv4a_3_3 = relu(batch_normalization(conv_2d(conv3b_1_1, 192, 3, bias=False, padding='VALID',activation=None, name='Conv2d_4a_3x3')))
maxpool5a_3_3 = max_pool_2d(conv4a_3_3, 3, strides=2, padding='VALID', name='MaxPool_5a_3x3')
tower_conv = relu(batch_normalization(conv_2d(maxpool5a_3_3, 96, 1, bias=False, activation=None, name='Conv2d_5b_b0_1x1')))
from tflearn.data_utils import image_preloader
#X, Y = image_preloader(files_list, image_shape=(224, 224), mode='file',
# categorical_labels=True, normalize=False,
# files_extension=['.jpg', '.png'], filter_channel=True)
# or use the mode 'floder'
# X, Y = image_preloader(data_dir, image_shape=(224, 224), mode='folder',
# categorical_labels=True, normalize=True,
# files_extension=['.jpg', '.png'], filter_channel=True)
num_classes = 17 # num of your dataset
datapath = '../../data/17flowers224'
import tflearn.datasets.oxflower17 as oxflower17
X, Y = oxflower17.load_data(one_hot=True, dirname=datapath)
# VGG preprocessing
img_prep = ImagePreprocessing()
img_prep.add_featurewise_zero_center(mean=[123.68, 116.779, 103.939],
per_channel=True)
# VGG Network
x = tflearn.input_data(shape=[None, 224, 224, 3], name='input',
data_preprocessing=img_prep)
softmax = vgg16(x, num_classes)
regression = tflearn.regression(softmax, optimizer='adam',
loss='categorical_crossentropy',
learning_rate=0.001, restore=False)
model = tflearn.DNN(regression, checkpoint_path='vnf-ckpt/vgg-finetuning-pre',
max_checkpoints=3, tensorboard_verbose=2,
# import dependencies
import tflearn.datasets.oxflower17 as oxflower17
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Conv2D, MaxPooling2D, Flatten, Dense
from tensorflow.keras.layers import BatchNormalization, Dropout
# load data
X, Y = oxflower17.load_data(one_hot=True,
dirname='data/17flowers')
# model specification
model = Sequential()
# first conv-pool block
model.add(Conv2D(96, kernel_size=(11, 11), strides=(4, 4), activation='relu', input_shape=(224, 224, 3)))
model.add(MaxPooling2D(pool_size=(3, 3), strides=(2, 2)))
model.add(BatchNormalization())
# second conv-pool block
model.add(Conv2D(256, kernel_size=(5, 5), activation='relu'))
model.add(MaxPooling2D(pool_size=(3, 3), strides=(2, 2)))
model.add(BatchNormalization())
# third conv-pool block
model.add(Conv2D(256, kernel_size=(3, 3), activation='relu'))
model.add(Conv2D(384, kernel_size=(3, 3), activation='relu'))
model.add(Conv2D(384, kernel_size=(3, 3), activation='relu'))
model.add(MaxPooling2D(pool_size=(3, 3), strides=(2, 2)))
model.add(BatchNormalization())
# dense block
# Day_03_01_alexnet_quiz.py import numpy as np import tensorflow as tf import tflearn.datasets.oxflower17 as oxflower17 # def load_data(dirname="17flowers_onehot", resize_pics=(224, 224), shuffle=True, one_hot=False): # 데이터셋 크기 : (1360, 224, 224, 3), (1360, 17) features, labels = oxflower17.load_data(one_hot=True) # print(features.shape, labels.shape) # print(features[:3]) # (1360, 224, 224, 3) (1360, 17) x_train, x_test = features[:1000], features[1000:] y_train, y_test = labels[:1000], labels[1000:] num_labels = 17 batch_size = 32 # [문제 1] shape을 채우세요. # 일단 여기서 None 대신 batch_size로 대신한다. # 학습할때만 batch size x = tf.placeholder(tf.float32, shape=(None, 224, 224, 3)) y = tf.placeholder(tf.float32, shape=(None, num_labels)) # ------------------------------------------------------------ # # 변수 생성 # [문제 2] 아래 숫자들을 채우세요. w1 = tf.Variable(tf.truncated_normal([11, 11, 3, 96], stddev=0.1)) # 필터 가로, 세로, RGB(채널) 11 x 11 x 3 피처갯수, 96 클래스갯수
Links:
- [GoogLeNet Paper](http://www.cv-foundation.org/openaccess/content_cvpr_2015/papers/Szegedy_Going_Deeper_With_2015_CVPR_paper.pdf)
- [Flower Dataset (17)](http://www.robots.ox.ac.uk/~vgg/data/flowers/17/)
"""
from __future__ import division, print_function, absolute_import
import tflearn
from tflearn.layers.core import input_data, dropout, fully_connected
from tflearn.layers.conv import conv_2d, max_pool_2d, avg_pool_2d
from tflearn.layers.normalization import local_response_normalization
from tflearn.layers.merge_ops import merge
from tflearn.layers.estimator import regression
import tflearn.datasets.oxflower17 as oxflower17
X, Y = oxflower17.load_data(one_hot=True, resize_pics=(227, 227))
network = input_data(shape=[None, 227, 227, 3])
conv1_7_7 = conv_2d(network, 64, 7, strides=2, activation='relu', name='conv1_7_7_s2')
pool1_3_3 = max_pool_2d(conv1_7_7, 3, strides=2)
pool1_3_3 = local_response_normalization(pool1_3_3)
conv2_3_3_reduce = conv_2d(pool1_3_3, 64, 1, activation='relu', name='conv2_3_3_reduce')
conv2_3_3 = conv_2d(conv2_3_3_reduce, 192, 3, activation='relu', name='conv2_3_3')
conv2_3_3 = local_response_normalization(conv2_3_3)
pool2_3_3 = max_pool_2d(conv2_3_3, kernel_size=3, strides=2, name='pool2_3_3_s2')
# 3a
inception_3a_1_1 = conv_2d(pool2_3_3, 64, 1, activation='relu', name='inception_3a_1_1')
inception_3a_3_3_reduce = conv_2d(pool2_3_3, 96, 1, activation='relu', name='inception_3a_3_3_reduce')
inception_3a_3_3 = conv_2d(inception_3a_3_3_reduce, 128, filter_size=3, activation='relu', name='inception_3a_3_3')
inception_3a_5_5_reduce = conv_2d(pool2_3_3, 16, filter_size=1, activation='relu', name='inception_3a_5_5_reduce')
Links:
- [GoogLeNet Paper](http://www.cv-foundation.org/openaccess/content_cvpr_2015/papers/Szegedy_Going_Deeper_With_2015_CVPR_paper.pdf)
- [Flower Dataset (17)](http://www.robots.ox.ac.uk/~vgg/data/flowers/17/)
"""
from __future__ import division, print_function, absolute_import
import tflearn
from tflearn.layers.core import input_data, dropout, fully_connected
from tflearn.layers.conv import conv_2d, max_pool_2d, avg_pool_2d
from tflearn.layers.normalization import local_response_normalization
from tflearn.layers.merge_ops import merge
from tflearn.layers.estimator import regression
import tflearn.datasets.oxflower17 as oxflower17
X, Y = oxflower17.load_data(one_hot=True, resize_pics=(227, 227))
network = input_data(shape=[None, 227, 227, 3])
conv1_7_7 = conv_2d(network,
64,
7,
strides=2,
activation='relu',
name='conv1_7_7_s2')
pool1_3_3 = max_pool_2d(conv1_7_7, 3, strides=2)
pool1_3_3 = local_response_normalization(pool1_3_3)
conv2_3_3_reduce = conv_2d(pool1_3_3,
64,
1,
activation='relu',
name='conv2_3_3_reduce')
def run(self):
hyperparams = self._parent.hyperparams
output_dir = hyperparams['output_dir']
epochs = int(hyperparams['epochs'])
batch_size = int(hyperparams['batch_size'])
validation_split = float(hyperparams['validation_split'])
conv2d_1_filters = int(hyperparams['conv2d_1_filters'])
conv2d_2_filters = int(hyperparams['conv2d_2_filters'])
conv2d_3_filters = int(hyperparams['conv2d_3_filters'])
conv2d_4_filters = int(hyperparams['conv2d_4_filters'])
conv2d_5_filters = int(hyperparams['conv2d_5_filters'])
dense_1_units = int(hyperparams['dense_1_units'])
dense_2_units = int(hyperparams['dense_2_units'])
dense_3_units = int(hyperparams['dense_3_units'])
drop_1_rate = float(hyperparams['drop_1_rate'])
drop_2_rate = float(hyperparams['drop_2_rate'])
fit_verbose = int(hyperparams['fit_verbose'])
X, Y = oxflower17.load_data(one_hot=True)
model = Sequential()
model.add(
Conv2D(conv2d_1_filters,
kernel_size=(11, 11),
strides=(4, 4),
activation='relu',
input_shape=(224, 224, 3)))
model.add(MaxPooling2D(pool_size=(3, 3), strides=(2, 2)))
model.add(BatchNormalization())
model.add(
Conv2D(conv2d_2_filters, kernel_size=(5, 5), activation='relu'))
model.add(MaxPooling2D(pool_size=(3, 3), strides=(2, 2)))
model.add(BatchNormalization())
model.add(
Conv2D(conv2d_3_filters, kernel_size=(3, 3), activation='relu'))
model.add(
Conv2D(conv2d_4_filters, kernel_size=(3, 3), activation='relu'))
model.add(
Conv2D(conv2d_5_filters, kernel_size=(3, 3), activation='relu'))
model.add(MaxPooling2D(pool_size=(3, 3), strides=(2, 2)))
model.add(BatchNormalization())
model.add(Flatten())
model.add(Dense(dense_1_units, activation='tanh'))
model.add(Dropout(drop_1_rate))
model.add(Dense(dense_2_units, activation='tanh'))
model.add(Dropout(drop_2_rate))
model.add(Dense(dense_3_units, activation='softmax'))
model.summary()
model.compile(loss='categorical_crossentropy',
optimizer='adam',
metrics=['accuracy'])
modelcheckpoint = ModelCheckpoint(filepath=output_dir +
"/weights.{epoch:02d}.hdf5")
if not os.path.exists(output_dir):
os.makedirs(output_dir)
tensorbrd = TensorBoard('tb_logs/alexnet')
self._parent.history = model.fit(
X,
Y,
batch_size=batch_size,
epochs=epochs,
verbose=fit_verbose,
validation_split=validation_split,
shuffle=True,
callbacks=[modelcheckpoint, tensorbrd])
weights_filepath = output_dir + "/weights.{:02d}.hdf5".format(
best_epoch(self._parent.history) + 1)
for cfg in model.get_config():
print(cfg)
print("epochs={}, batch_size={} best weights filepath {}".format(
epochs, batch_size, weights_filepath))
evt = TrainingDoneEvent(EVT_WORK_DONE_TYPE, -1)
wx.PostEvent(self._parent, evt)
def download():
oxflower17.load_data('Total_Data/flowers17/')
import keras
import tensorflow as tf
import tflearn
from keras.models import Sequential
from keras.layers import Dense, Activation, Dropout, Flatten,\
Conv2D, MaxPooling2D
from keras.layers.normalization import BatchNormalization
import numpy as np
np.random.seed(1000)
# (2) Get Data
import tflearn.datasets.oxflower17 as oxflower17
x, y = oxflower17.load_data(one_hot=True)
# (3) Create a sequential model
model = Sequential()
# 1st Convolutional Layer
model.add(Conv2D(filters=96, input_shape=(224,224,3), kernel_size=(11,11),\
strides=(4,4), padding='valid'))
model.add(Activation('relu'))
# Pooling
model.add(MaxPooling2D(pool_size=(2,2), strides=(2,2), padding='valid'))
# Batch Normalisation before passing it to the next layer
model.add(BatchNormalization())
# 2nd Convolutional Layer
model.add(Conv2D(filters=256, kernel_size=(11,11), strides=(1,1), padding='valid'))
model.add(Activation('relu'))
# Pooling
model.add(MaxPooling2D(pool_size=(2,2), strides=(2,2), padding='valid'))
# 全连接层1
network = fully_connected(network, 4096, activation='tanh')
network = dropout(network, 0.5)
# 全连接层2
network = fully_connected(network, 4096, activation='tanh')
network = dropout(network, 0.5)
# 输出层
network = fully_connected(network, 17, activation='softmax')
network = regression(network, optimizer='momentum', loss='categorical_crossentropy', learning_rate=0.001)
return network
if __name__ == '__main__':
# 加载数据
dataset = 'alexnet_oxflower17'
X, Y = oxflower17.load_data(dirname='../datasets/17flowers', one_hot=True, resize_pics=(227, 227))
# 构建模型
alexnet = alex_net()
modal = tflearn.DNN(alexnet, checkpoint_path='./model/', max_checkpoints=1, tensorboard_verbose=2)
# 检查点
model_file = './model/' + dataset + '.model'
if os.path.isfile(model_file):
modal.load(model_file)
try:
modal.fit(X, Y, n_epoch=10, validation_set=0.2, shuffle=True,
show_metric=True, batch_size=16, snapshot_step=200,
snapshot_epoch=True, run_id=dataset)
modal.save(model_file)
except KeyboardInterrupt as i:
print('Closed by an KeyboardInterrupt')
finally:
Links:
[MNIST Dataset] http://yann.lecun.com/exdb/mnist/
"""
from __future__ import division, print_function, absolute_import
import numpy as np
import matplotlib.pyplot as plt
import tflearn
import tflearn.datasets.oxflower17 as oxflower17
import time
# Data loading and preprocessing
import tflearn.datasets.mnist as mnist
X, Y = oxflower17.load_data(one_hot=True, resize_pics=(28, 28))
X = np.reshape(X, (len(X), 2352))
testX = X[1000:]
X = X[:1000]
start_time = time.time()
# Building the encoder
encoder = tflearn.input_data(shape=[None, 2352])
encoder = tflearn.fully_connected(encoder, 2100)
encoder = tflearn.fully_connected(encoder, 200)
# Building the decoder
decoder = tflearn.fully_connected(encoder, 2100)
decoder = tflearn.fully_connected(decoder, 2352, activation='sigmoid')
# Regression, with mean square error
(X, Y), (testX, testY) = cifar10.load_data()
X, Y = shuffle(X, Y)
Y = to_categorical(Y)
testY = to_categorical(testY)
X = shapeToOneD(X)
Y = shapeToOneD(Y)
testX = shapeToOneD(testX)
testY = shapeToOneD(testY)
elif jdata.model == "cifar100": # input 1024 - output 100
print("https://www.cs.toronto.edu/~kriz/cifar.html")
from tflearn.datasets import cifar100
(X, Y), (testX, testY) = cifar100.load_data()
elif jdata.model == "oxflower17.py": # input 50176 - output 17
print("http://www.robots.ox.ac.uk/~vgg/data/flowers/17/")
from tflearn.datasets import oxflower17
(X, Y) = oxflower17.load_data()
elif jdata.model == "svhn": # input 1024 - output 10
print("http://ufldl.stanford.edu/housenumbers")
from tflearn.datasets import svhn
X, Y, testX, testY = svhn.load_data()
else:
sys.exit(1)
# Building deep neural network
net = tflearn.input_data(shape=[None, inputlayer], name='input')
modelFilename = ""
for i in range(innerLayers):
modelFilename += str(jdata.layers[i].nodes) + str(jdata.layers[i].activation)[:5]
net = tflearn.fully_connected(net, int(jdata.layers[i].nodes), bias=jdata.layers[i].bias, activation=str(jdata.layers[i].activation), name='dense'+str(i))
# output layer
dtype=tf.float32,
initializer=tf.random_normal_initializer(mean=0.0,
stddev=0.1)):
"""
定义获取w,b随机初始值得函数封装，w值对应卷积的卷积窗口大小
:param name: 变量的命名值
:param shape: 变量的形状
:param dtype: 变量的数据类型
:param initializer: 变量的初始化
:return: tensor的初始化对象
"""
return tf.get_variable(name, shape, dtype, initializer)
# 一 加载数据
X, Y = oxflower17.load_data(dirname="17flowers", one_hot=True)
print(X.shape) # (1360,224, 224, 3)
print(Y.shape) # (1360,17)
# 二 构建网络
# 1.超参数
# 2.变量类型w,b,占位符
x_input = tf.placeholder(shape=[None, 224, 224, 3], dtype=tf.float32)
y_out = tf.placeholder(shape=[None, 17], dtype=tf.float32)
learn_rate = tf.placeholder(dtype=tf.float32)
# 3.VGG网络
def VGG():
# 第一层卷积层(224,224,64)
