To specify the decoder model architecture, we will use the following code:

# Decoder
decoder <- encoder %>% 
         layer_conv_2d(filters = 4, 
                       kernel_size = c(3,3), 
                       activation = 'relu',
                       padding = 'same') %>%   
         layer_upsampling_2d(c(2,2)) %>% 
         layer_conv_2d(filters = 8, 
                       kernel_size = c(3,3), 
                       activation = 'relu',
                       padding = 'same') %>%  
         layer_upsampling_2d(c(2,2)) %>% 
         layer_conv_2d(filters = 1, 
                       kernel_size = c(3,3), 
                       activation = 'sigmoid',
                       padding = 'same')
summary(decoder)
Output
Tensor("conv2d_25/Sigmoid:0", shape=(?, 28, 28, 1), dtype=float32)

Here, the encoder model has become the input for the decoder model. For the decoder network, we use a similar structure, with the first convolutional layer having 4 filters and the second convolutional layer having 8 filters. In addition, instead of pooling layers, we now use up-sampling layers. The first upsampling layer changes the height and width to 14 x 14 and the second upsampling layer restores it to the original height and width of 28 x 28. In the last layer, we make use of the sigmoid activation function, which ensures that the output values remain between 0 and 1.