The autoencoder model and the summary of the model showing the output shape and the number of parameters for each layer is as follows:

# Autoencoder
ae_model <- keras_model(inputs = input_layer, outputs = decoder)
summary(ae_model)
__________________________________________________________________________
Layer (type)                      Output Shape               Param #       
==========================================================================
input_5 (InputLayer)              (None, 28, 28, 1)            0             
__________________________________________________________________________
conv2d_21 (Conv2D)                (None, 28, 28, 8)            80            
__________________________________________________________________________
max_pooling2d_9 (MaxPooling2D)    (None, 14, 14, 8)             0             
__________________________________________________________________________
conv2d_22 (Conv2D)                (None, 14, 14, 4)            292           
__________________________________________________________________________
max_pooling2d_10 (MaxPooling2D)   (None, 7, 7, 4)               0             
__________________________________________________________________________
conv2d_23 (Conv2D)                (None, 7, 7, 4)              148           
___________________________________________________________________________
up_sampling2d_9 (UpSampling2D)    (None, 14, 14, 4)             0             
___________________________________________________________________________
conv2d_24 (Conv2D)                (None, 14, 14, 8)            296           
___________________________________________________________________________
up_sampling2d_10 (UpSampling2D)   (None, 28, 28, 8)             0             
___________________________________________________________________________
conv2d_25 (Conv2D)                (None, 28, 28, 1)             73           
===========================================================================
Total params: 889
Trainable params: 889
Non-trainable params: 0
____________________________________________________________________________________

Here, the autoencoder model has five convolutional layers, two maximum pooling layers, and two upsampling layers, apart from the input layer. Here, the total number of parameters in this autoencoder model is 889.