What are convolutional autoencoders?

They use convolutional layers instead of fully connected layers for image data.

Convolutional autoencoders capture spatial hierarchies in images using convolution and pooling layers in encoder and upsampling or transposed convolutions in decoder.