Skip to content

Instantly share code, notes, and snippets.

@iCorv

iCorv/subpixel1d_cnn.py

Last active March 15, 2021 13:47
Show Gist options
  • Star 0 You must be signed in to star a gist
  • Fork 0 You must be signed in to fork a gist
  • Save iCorv/99ce9366b013a1e948b2fe757c340699 to your computer and use it in GitHub Desktop.
Save iCorv/99ce9366b013a1e948b2fe757c340699 to your computer and use it in GitHub Desktop.
Download ZIP
A simple auto-encoder using the Subpixel1D layer for upsampling
Raw
subpixel1d_cnn.py
# down- and up-sampling by a factor of 4
strides = 4
inputs = tf.keras.Input(shape=(16384, 1))
d = tf.keras.layers.Conv1D(16, kernel_size=64, strides=strides,
padding='same', activation='elu',
kernel_initializer='he_normal')(inputs)
d = tf.keras.layers.Conv1D(32, kernel_size=32, strides=strides, padding='same',
activation='elu', kernel_initializer='he_normal')(d)
d = tf.keras.layers.Conv1D(64, kernel_size=16, strides=strides, padding='same',
activation='elu', kernel_initializer='he_normal')(d)
bottleneck = tf.keras.layers.Conv1D(32, kernel_size=8,
strides=1, padding='same',
activation='elu',
kernel_initializer='he_normal',
name='bottleneck')(d)
u = tf.keras.layers.Conv1D(64, kernel_size=16, strides=1,
padding='same', activation='elu',
kernel_initializer='he_normal')(bottleneck)
u = Subpixel1D(r=strides)(u)
u = tf.keras.layers.Conv1D(32, kernel_size=32, strides=1, padding='same',
activation='elu', kernel_initializer='he_normal')(u)
u = Subpixel1D(r=strides)(u)
u = tf.keras.layers.Conv1D(strides, kernel_size=64, strides=1, padding='same',
activation='elu', kernel_initializer='he_normal')(u)
outputs = Subpixel1D(r=strides)(u)
model = tf.keras.Model(inputs=inputs, outputs=outputs)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment