Dynamic 3D Gaussians: Tracking by Persistent Dynamic View Synthesis

TL;DR

This paper introduces Dynamic 3D Gaussians, a novel method for simultaneous dynamic scene view synthesis and 6-DOF tracking, enabling applications like 4D video editing without explicit correspondence inputs.

Contribution

It proposes a new dynamic scene representation with persistent Gaussians that move and rotate over time, improving dynamic reconstruction and tracking capabilities.

Findings

Accurately models dynamic scenes with persistent Gaussians.

Enables 6-DOF tracking without correspondence or flow inputs.

Supports diverse applications like view synthesis and video editing.

Abstract

We present a method that simultaneously addresses the tasks of dynamic scene novel-view synthesis and six degree-of-freedom (6-DOF) tracking of all dense scene elements. We follow an analysis-by-synthesis framework, inspired by recent work that models scenes as a collection of 3D Gaussians which are optimized to reconstruct input images via differentiable rendering. To model dynamic scenes, we allow Gaussians to move and rotate over time while enforcing that they have persistent color, opacity, and size. By regularizing Gaussians' motion and rotation with local-rigidity constraints, we show that our Dynamic 3D Gaussians correctly model the same area of physical space over time, including the rotation of that space. Dense 6-DOF tracking and dynamic reconstruction emerges naturally from persistent dynamic view synthesis, without requiring any correspondence or flow as input. We demonstrate a large number of downstream applications enabled by our representation, including first-person view synthesis, dynamic compositional scene synthesis, and 4D video editing.