def inference_AlexNet(self,im):
l1 = self._AlexNet_conv_layer(im, 96, (11,11), (4,4), (3,3), (2,2), name='l1')
l2 = self._AlexNet_conv_layer(l1, 256, (5,5), (1,1), (3,3), (2,2), name='l2')
with tf.variable_scope('l3') as scope:
l3_conv = tl.conv2d(l2, 384, (3,3), (1,1), name='l3')
l3 = tl.relu(l3_conv)
with tf.variable_scope('l4') as scope:
l4_conv = tl.conv2d(l3, 384, (3,3), (1,1), name='l4')
l4 = tl.relu(l4_conv)
with tf.variable_scope('l5') as scope:
l5_lconv = tl.conv2d(l4, 256, (3,3), (1,1), name='l5')
l5_lrelu = tl.relu(l5_lconv)
l5 = tl.max_pool2d(l5_lrelu, (3,3), (2,2), padding='VALID', name='pool')
#print l5.get_shape()
fc0 = tf.reshape(l5, [self.batch_size, -1])
fc1 = tl.fc(fc0, 4096, name='fc1')
l6_lrelu = tl.relu(fc1)
l6 = tl.dropout(l6_lrelu, self.keep_prob)
fc2 = tl.fc(l6, 4096, name='fc2')
l7_lrelu = tl.relu(fc2)
l7 = tl.dropout(l7_lrelu, self.keep_prob)
fc3 = tl.fc(l7, 10, name='fc3')
return fc3
def inference(self, im):
l0 = im
l1 = self._conv_layer(l0, 32, (5, 5), (1, 1), (2, 2), (2, 2), name='l1')
l2 = self._conv_layer(l1, 32, (5, 5), (1, 1), (2, 2), (2, 2), name='l2')
l3 = self._conv_layer(l2, 64, (5, 5), (1, 1), (2, 2), (2, 2), name='l3')
fc0 = tf.reshape(l3, [self.batch_size, -1])
fc1 = tl.fc(fc0, 64, name='fc1')
fc2 = tl.fc(fc1, 10, name='fc2')
return fc2
def inference_VGG19(self, im):
with tf.variable_scope('l1_conv1') as scope:
l1_conv1 = tl.conv2d(im, 64, (3, 3), (1, 1), name='l1')
with tf.variable_scope('l2_conv2') as scope:
l2_conv2 = tl.conv2d(l1_conv1, 64, (3, 3), (1, 1), name='l2')
l2_pooling = tl.max_pool2d(l2_conv2, (3, 3), (2, 2), padding='SAME', name='pool')
with tf.variable_scope('l3_conv1') as scope:
l3_conv1 = tl.conv2d(l2_pooling, 128, (3, 3), (1, 1), name='l3')
with tf.variable_scope('l4_conv2') as scope:
l4_conv2 = tl.conv2d(l3_conv1, 128, (3, 3), (1, 1), name='l4')
l4_pooling = tl.max_pool2d(l4_conv2, (3, 3), (2, 2), padding='SAME', name='pool')
with tf.variable_scope('l5_conv1') as scope:
l5_conv1 = tl.conv2d(l4_pooling, 256, (3, 3), (1, 1), name='l5')
with tf.variable_scope('l6_conv2') as scope:
l6_conv2 = tl.conv2d(l5_conv1, 256, (3, 3), (1, 1), name='l6')
with tf.variable_scope('l7_conv3') as scope:
l7_conv3 = tl.conv2d(l6_conv2, 256, (3, 3), (1, 1), name='l7')
with tf.variable_scope('l8_conv4') as scope:
l8_conv4 = tl.conv2d(l7_conv3, 256, (3, 3), (1, 1), name='l8')
l8_pooling = tl.max_pool2d(l8_conv4, (3, 3), (2, 2), padding='SAME', name='pool')
with tf.variable_scope('l9_conv1') as scope:
l9_conv1 = tl.conv2d(l8_pooling, 512, (3, 3), (1, 1), name='l9')
with tf.variable_scope('l10_conv2') as scope:
l10_conv2 = tl.conv2d(l9_conv1, 512, (3, 3), (1, 1), name='l10')
with tf.variable_scope('l11_conv3') as scope:
l11_conv3 = tl.conv2d(l10_conv2, 512, (3, 3), (1, 1), name='l11')
with tf.variable_scope('l12_conv4') as scope:
l12_conv4 = tl.conv2d(l11_conv3, 512, (3, 3), (1, 1), name='l12')
l12_pooling = tl.max_pool2d(l12_conv4, (3, 3), (2, 2), padding='SAME', name='pool')
with tf.variable_scope('l13_conv1') as scope:
l3_conv1 = tl.conv2d(l12_pooling, 512, (3, 3), (1, 1), name='l13')
with tf.variable_scope('l14_conv2') as scope:
l14_conv2 = tl.conv2d(l3_conv1, 512, (3, 3), (1, 1), name='l14')
with tf.variable_scope('l15_conv3') as scope:
l15_conv3 = tl.conv2d(l14_conv2, 512, (3, 3), (1, 1), name='l15')
with tf.variable_scope('l16_conv4') as scope:
l16_conv4 = tl.conv2d(l15_conv3, 512, (3, 3), (1, 1), name='l16')
l16_pooling = tl.max_pool2d(l16_conv4, (3, 3), (2, 2), padding='SAME', name='pool')
fc0 = tf.reshape(l16_pooling, [self.batch_size, -1])
fc1 = tl.fc(fc0, 4096, name='fc1')
fc1_out = tl.dropout(fc1, self.keep_prob)
fc2 = tl.fc(fc1_out, 4096, name='fc2')
fc2_out = tl.dropout(fc2, self.keep_prob
)
fc3 = tl.fc(fc2_out, 10, name='fc3')
return fc3
def inference_MobileNet(self, im):
n=6
num_classes=10
width_multiplier=1
net = tl.conv2d(im, round(32 * width_multiplier), (3, 3), (2, 2), padding='SAME', name='conv_1')
net = tl.bn_new(net, name='conv_1')
net = tl.relu(net)
net = self.MobileNet_separable_2d(net, 32, name ='sep1')
net = self.MobileNet_separable_2d(net, 64, (2, 2), name='sep2')
net = self.MobileNet_separable_2d(net, 128, name ='sep3')
net = self.MobileNet_separable_2d(net, 128, (2, 2), name='sep4')
net = self.MobileNet_separable_2d(net, 256, name ='sep5')
net = self.MobileNet_separable_2d(net, 256, (2, 2), name='sep6')
net = self.MobileNet_separable_2d(net, 512, name ='sep7')
for i in range(n):
net = self.MobileNet_separable_2d(net, 512, name ='sep' + str(i + 8))
net = self.MobileNet_separable_2d(net, 512, (2, 2), name ='sep' + str(n + 9))
net = self.MobileNet_separable_2d(net, 1024, name='sep' + str(n + 10))
#shape = net.get_shape()
#net = tl.avg_pool2d(net, shape[1:3], name='avg_pool')
net = tf.reshape(net, [self.batch_size, -1])
logits = tl.fc(net, num_classes, name='fc')
return logits
def inference_ResNet(self, im):
n=5
n_class=10
with tf.variable_scope('conv_0') as scope:
y = tl.conv2d(im, 16, (3,3), (1,1), padding='SAME', name='conv_init')
y = tl.bn_new(y)
y = tl.relu(y)
y = self.residual_group(y, 16, n, False, name='group_1')
y = self.residual_group(y, 32, n, True, name='group_2')
y = self.residual_group(y, 64, n, True, name='group_3')
#print(y.get_shape())
#shape = y.get_shape()
#y = tl.avg_pool2d(y, shape[1:3], (1, 1), padding='VALID', name='avg_pool')
y = tf.reshape(y, [self.batch_size, -1])
y = tl.fc(y, n_class, name='fc')
return y
def inference_Inceptionv3(self,im):
n=3
# n_class=10
n_class=101
with tf.variable_scope('Inceptionv3') as scope:
y = tl.conv2d(im, 32, (3, 3), (2, 2), name='conv0')
y = tl.bn_new(y,name='conv1')
y = tl.relu(y)
y = tl.conv2d(y, 32, (3, 3), (1, 1), name='conv1')
y = tl.bn_new(y,name='conv2')
y = tl.relu(y)
y = tl.conv2d(y, 64, (3, 3), (1, 1), padding='SAME', name='conv2')
y = tl.bn_new(y,name='conv3')
y = tl.relu(y)
y = tl.max_pool2d(y, (3, 3), (2, 2), name='pool1')
y = tl.bn_new(y,name='conv4')
y = tl.relu(y)
# 73 x 73 x 64
y = tl.conv2d(y, 80, (3, 3), (1, 1), name='conv3')
y = tl.bn_new(y,name='conv5')
y = tl.relu(y)
# 73 x 73 x 80.
y = tl.conv2d(y, 192, (3, 3), (2, 2), name='conv4')
y = tl.bn_new(y,name='conv6')
y = tl.relu(y)
# 71 x 71 x 192.
y = tl.max_pool2d(y, (3, 3), (2, 2), name='pool2')
# 35 x 35 x 192.
for i in range(n - 1):
y = self.Inceptionv3_module1(y, (5, 5),name='mixed_35x35x256a'+str(i))
for i in range(n - 1):
y = self.Inceptionv3_module2(y,name='mixed_17x17x768e'+str(i))
for i in range(n - 1):
y = self.Inceptionv3_module3(y,name='mixed_8x8x2048a'+str(i))
shape = y.get_shape()
y = tl.avg_pool2d(y, shape[1:3], padding='VALID', name='pool')
# 1 x 1 x 2048
y = tl.dropout(y, self.keep_prob)
y = tl.flatten(y)
# 2048
logits = tl.fc(y, n_class, name='logits')
return logits