Last active
December 1, 2019 05:00
-
-
Save luis-gonzales/464480f08e3df7a4bd98a0734b266737 to your computer and use it in GitHub Desktop.
Keras layer for MobileNetV2
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| import tensorflow as tf | |
| from tensorflow.keras.layers import Layer, Conv2D, DepthwiseConv2D, BatchNormalization | |
| class InvertedResidual(Layer): | |
| def __init__(self, filters, strides, expansion_factor=6, trainable=True, | |
| name=None, **kwargs): | |
| super(InvertedResidual, self).__init__(trainable=trainable, name=name, **kwargs) | |
| self.filters = filters | |
| self.strides = strides | |
| self.expansion_factor = expansion_factor # allowed to be decimal value | |
| def build(self, input_shape): | |
| input_channels = int(input_shape[3]) | |
| self.ptwise_conv1 = Conv2D(filters=int(input_channels*self.expansion_factor), | |
| kernel_size=1, use_bias=False) | |
| self.dwise = DepthwiseConv2D(kernel_size=3, strides=self.strides, | |
| padding='same', use_bias=False) | |
| self.ptwise_conv2 = Conv2D(filters=self.filters, kernel_size=1, use_bias=False) | |
| self.bn1 = BatchNormalization() | |
| self.bn2 = BatchNormalization() | |
| self.bn3 = BatchNormalization() | |
| def call(self, input_x): | |
| # Expansion to high-dimensional space | |
| x = self.ptwise_conv1(input_x) | |
| x = self.bn1(x) | |
| x = tf.nn.relu6(x) | |
| # Spatial filtering | |
| x = self.dwise(x) | |
| x = self.bn2(x) | |
| x = tf.nn.relu6(x) | |
| # Projection back to low-dimensional space w/ linear activation | |
| x = self.ptwise_conv2(x) | |
| x = self.bn3(x) | |
| # Residual connection if i/o have same spatial and depth dims | |
| if input_x.shape[1:] == x.shape[1:]: | |
| x += input_x | |
| return x | |
| def get_config(self): | |
| cfg = super(InvertedResidual, self).get_config() | |
| cfg.update({'filters': self.filters, | |
| 'strides': self.strides, | |
| 'expansion_factor': self.expansion_factor}) | |
| return cfg |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
thank you so much!