Python Deep Learning Projects

Additional topic – convolution autoencoder

An autoencoder is a combination of two parts: an encoder and a decoder. The encoder and decoder of a simple autoencoder are usually made up of dense layers, whereas in a convolution autoencoder, they are made of convolution layers:

Figure 8.37: The structure of an autoencoder (image source: Wikipedia)

The encoder part of the autoencoder accepts an image and compresses it into a smaller size with the help of a pooling operation. In our case, this is max pooling. The decoder accepts the input of the encoder and learns to expand the image to our desired size by using convolution and upsampling.

Imagine ...