ImplicitDeepfake: Plausible Face-Swapping through Implicit Deepfake Generation using NeRF and Gaussian Splatting

Numerous emerging deep-learning techniques have had a substantial impact on computer graphics. Among the most promising breakthroughs are the recent rise of Neural Radiance Fields (NeRFs) and Gaussian Splatting (GS). NeRFs encode the object's shape and color in neural network weights using a handful of images with known camera positions to generate novel views. In contrast, GS provides accelerated training and inference without a decrease in rendering quality by encoding the object's characteristics in a collection of Gaussian distributions. These …

color computer computer graphics cs.cv deepfake encode face fields graphics images impact nerf network neural network neural radiance fields through