def mobilenetv2_yolo_body(inputs, num_anchors, num_classes, alpha=1.0):
mobilenetv2 = mobilenet_v2(default_batchnorm_momentum=0.9,
alpha=alpha,
input_tensor=inputs,
include_top=False,
weights='imagenet')
x, y1 = make_last_layers_mobilenet(mobilenetv2.output, 17, 512,
num_anchors * (num_classes + 5))
x = compose(
tf.keras.layers.Conv2D(256,
kernel_size=1,
padding='same',
use_bias=False,
name='block_20_conv'),
tf.keras.layers.BatchNormalization(momentum=0.9, name='block_20_BN'),
tf.keras.layers.ReLU(6., name='block_20_relu6'),
tf.keras.layers.UpSampling2D(2))(x)
x = tf.keras.layers.Concatenate()([
x,
MobilenetConv2D(
(1, 1), alpha,
384)(mobilenetv2.get_layer('block_12_project_BN').output)
])
x, y2 = make_last_layers_mobilenet(x, 21, 256,
num_anchors * (num_classes + 5))
x = compose(
tf.keras.layers.Conv2D(128,
kernel_size=1,
padding='same',
use_bias=False,
name='block_24_conv'),
tf.keras.layers.BatchNormalization(momentum=0.9, name='block_24_BN'),
tf.keras.layers.ReLU(6., name='block_24_relu6'),
tf.keras.layers.UpSampling2D(2))(x)
x = tf.keras.layers.Concatenate()([
x,
MobilenetConv2D(
(1, 1), alpha,
128)(mobilenetv2.get_layer('block_5_project_BN').output)
])
x, y3 = make_last_layers_mobilenet(x, 25, 128,
num_anchors * (num_classes + 5))
y1 = tf.keras.layers.Lambda(lambda y: tf.reshape(y, [
-1, tf.shape(y)[1],
tf.shape(y)[2], num_anchors, num_classes + 5
]),
name='y1')(y1)
y2 = tf.keras.layers.Lambda(lambda y: tf.reshape(y, [
-1, tf.shape(y)[1],
tf.shape(y)[2], num_anchors, num_classes + 5
]),
name='y2')(y2)
y3 = tf.keras.layers.Lambda(lambda y: tf.reshape(y, [
-1, tf.shape(y)[1],
tf.shape(y)[2], num_anchors, num_classes + 5
]),
name='y3')(y3)
return AdvLossModel(mobilenetv2.inputs, [y1, y2, y3])
def mobilenetv2(inputs, alpha, classes):
"""MobilenetV2 wrapper function
Arguments:
inputs {np.array} -- [train images]
alpha {float} -- [controls the width of the network. This is known as the
width multiplier in the MobileNetV2 paper.
- If `alpha` < 1.0, proportionally decreases the number
of filters in each layer.
- If `alpha` > 1.0, proportionally increases the number
of filters in each layer.
- If `alpha` = 1, default number of filters from the paper
are used at each layer.]
classes {int} -- [classes total number]
Returns:
[tf.keras.Model] -- [mobilenetv2 model]
"""
return mobilenet_v2(default_batchnorm_momentum=0.9,
alpha=alpha,
input_tensor=inputs,
classes=classes)
def mobilenetv2(inputs, alpha, classes):
return mobilenet_v2(default_batchnorm_momentum=0.9,
alpha=alpha,
input_tensor=inputs,
classes=classes)