def train(x_train, x_test, y_size, x_size, num_channels, latent_space_dim,
learning_rate):
autoencoder, encoder, decoder, encoder_mu, encoder_log_variance = VAE(
y_size, x_size, num_channels, latent_space_dim)
autoencoder.summary()
autoencoder.compile(optimizer=Adam(lr=learning_rate),
loss=loss_func(encoder_mu, encoder_log_variance))
return autoencoder, encoder, decoder
def train(x_train, learning_rate, batch_size, epochs):
autoencoder = VAE(input_shape=(28, 28, 1),
conv_filters=(32, 64, 64, 64),
conv_kernels=(3, 3, 3, 3),
conv_strides=(1, 2, 2, 1),
latent_space_dim=100)
autoencoder.summary()
autoencoder.compile(learning_rate)
autoencoder.train(x_train, batch_size, epochs)
return autoencoder
def train(x_train, learning_rate, batch_size, epochs):
autoencoder = VAE(
input_shape=(256, 388, 1),
conv_filters=(512, 256), #(512, 256, 128, 64, 32)
conv_kernels=(3, 3), # (3, 3, 3, 3, 3)
conv_strides=(2, 2), # (2, 2, 2, 2, (2, 1))
latent_space_dim=128)
autoencoder.summary()
autoencoder.compile(learning_rate)
autoencoder.train(x_train, batch_size, epochs)
return autoencoder