Input: multi-view videoFrame 1Frame 100 Frame 200Output: animatable human model Reposed geometries Synthesized imagesFigure 1: Given a multi-view video of a performer (three views from Human3.6M [18] in this example), we reconstruct an animatable human model, which can be used for free view synthesis and 3D shape generation under novel human poses.

Input videoFrame 100 Frame 110 Frame 120 … …Animatable NeRF AniSDF (Ours)Figure 1. Given a monocular video of a performer, our method reconstructs high-quality geometries and renders photorealistic images under novel human poses, which outperforms Animatable NeRF [56]. More results can be found at https://zju3dv.github.io/animatable sdf/.

No abstract

No abstract