TL;DR通过无监督学习方法,提出了一种基于分解动态物体运动和相机运动的动态视图合成方法,包括非监督表面一致性和基于路径的多视图约束,可实现准确的 3D 运动建模,从而提高合成质量并相比现有方法精确地生成场景流和深度
Abstract
The challenge of dynamic view synthesis from dynamic monocular videos, i.e.,
synthesizing novel views for free viewpoints given a monocular video of a
dynamic scene captured by a moving camera, mainly lies in acc