Replacement using autoencoders
Deepfakes are an interesting and powerful use of technology that is both useful and dangerous. In previous sections, we discussed different modes of operations and key features that can be leveraged, as well as common architectures. We also briefly touched upon the high-level flow of different tasks required to achieve the end results. In this section, we will focus on developing a face swapping setup using an autoencoder as our backbone architecture. Let's get started.
Task definition
The aim of this exercise is to develop a face swapping setup. As discussed earlier, face swapping is a type of replacement mode operation in the context of deepfake terminology. In this setup, we will focus on transforming Nicolas Cage (a Hollywood actor) into Donald J. Trump (former US president). In the upcoming sections, we will present each sub-task necessary for the preparation of data, training our models, and finally, the generation of swapped fake...
