GoMAvatar, a novel approach for real-time, memory-efficient, high-quality animatable human modeling from a single monocular video, using a hybrid Gaussians-on-Mesh (GoM) representation.
