RigPAPR: Rig-Based Animation of Static Neural Point Clouds from a Fixed-Viewpoint Video
Researchers have developed RigPAPR, a novel method for animating static neural point clouds from fixed-viewpoint videos. Unlike previous techniques that suffer from joint-boundary artifacts due to rigid deformations, RigPAPR reconstructs surfaces at render time, allowing for natural articulation. This approach automatically rigs a point cloud and drives it using direct linear blend skinning, achieving superior rendering quality and accuracy compared to existing methods on synthetic and real-world data. AI
IMPACT Enables more realistic and artifact-free animation of 3D assets from video, potentially impacting CGI and virtual reality.