Dynamic View Synthesis from Dynamic Monocular Video

Dynamic View Synthesis from Dynamic Monocular Video

May 14, 2021
|
69 views
|
Details
Abstract: We present an algorithm for generating novel views at arbitrary viewpoints and any input time step given a monocular video of a dynamic scene. Our work builds upon recent advances in neural implicit representation and uses continuous and differentiable functions for modeling the time-varying structure and the appearance of the scene. We jointly train a time-invariant static NeRF and a time-varying dynamic NeRF, and learn how to blend the results in an unsupervised manner. However, learning this implicit function from a single video is highly ill-posed (with infinitely many solutions that match the input video). To resolve the ambiguity, we introduce regularization losses to encourage a more physically plausible solution. We show extensive quantitative and qualitative results of dynamic view synthesis from casually captured videos. Authors: Chen Gao, Ayush Saraf, Johannes Kopf, Jia-Bin Huang (Virginia Tech, Facebook)

Comments
loading...