Tianshu Kuai
Akash
Description: CAMM: Building Category-Agnostic and Animatable 3D Models from Monocular Videos
3d reconstruction (45) category-agnostic (1) pose manipulation (1) monocular videos (1)
CVPR 2023 DynaVis Workshop
Given monocular videos of an articulated object, our method builds its canonical representation, including 3D shape, appearance, and a corresponding animatable 3D kinematic chain for direct pose manipulations. Our approach does not rely on any information on the object's shape and underlying structure. In c) we show example of re-posing the human in b) to novel pose. We show the learned canonical representations of other object categories in d).
Animating an object in 3D often requires an articulated structure, e.g. a kinematic chain or skeleton of the manipulated object with proper skinning weights, to obtain smooth movements and surface deformations. However, existing models that allow direct pose manipulations are either limited to specific object categories or built with specialized equipment. To reduce the work needed for creating animatable 3D models, we propose a novel reconstruction method that learns an animatable kinematic chain for any a
Tianshu Kuai
Akash