Summary
Deepfakes are a complicated subject both ethically and technically. In this chapter, we discussed the deepfake technology in general to start with. We presented an overview of what deepfakes are all about and briefly touched upon a number of productive as well as malicious use cases. We presented a detailed discussion on different modes of operation of different deepfake setups and how each of these impacts the overall believability of generated content. While deepfakes is an all-encompassing term associated with videos, images, audio, text, and so on, we focused on visual use cases only in this chapter.
Given our scope, we discussed various feature sets leveraged by different works in this space. Particularly, we discussed the Facial Action Coding System (FACS), 3D morphable models (3DMM), and facial landmarks. We also discussed how we can perform facial landmark detection using libraries such as dlib and MTCNN. We then presented a high-level flow of tasks to be performed...
