With recent developments in AI, it is now possible to create high quality fake videos that look extremely realistic. This has positive applications in computer graphics, but on the other hand this can also have dangerous implications on society as in political propaganda and for public shaming. Even if we have a reliable detector(classifier) for one forgery, it is still unsure that it will work on a different forgery. This works aims to address this problem, transferability among different forgeries. My talk will focus on how to build a single model that detects most of the forgeries surfacing the internet.
Shivangi Aneja (TUM)