Applications of autoencoders

In the previous example of constructing images from a lower representation, we saw it was very similar to the original input, and also we saw the benefits of CANs while denoising the noisy dataset. This kind of example we have implemented above is really useful for the image construction applications and dataset denoising. So you can generalize the above implementation to any other example of interest to you.

Also, throughout this chapter, we have seen how flexible the autoencoder architecture is and how we can make different changes to it. We have even tested it to solve harder problems of removing noise from input images. This kind of flexibility opens the door to many more applications that auoencoders will be a great fit for.

Image colorization

Autoencoders—especially the convolutional version—can be used for harder tasks such as image colorization. In the following example, we feed the model with an input image without any colors, and the reconstructed version...