Dynamic 2D Gaussians: Geometrically Accurate Radiance Fields for Dynamic Objects

TL;DR

This paper introduces Dynamic 2D Gaussians (D-2DGS), a novel method for reconstructing accurate, high-quality meshes of dynamic objects from sparse images, outperforming existing 4D representations in mesh quality.

Contribution

The paper presents a new representation using 2D Gaussians and sparse-controlled points to achieve geometrically accurate dynamic mesh reconstruction from limited input data.

Findings

Outperforms existing methods in mesh detail and smoothness

Successfully reconstructs dynamic meshes from sparse images

Removes floaters for cleaner reconstructions

Abstract

Reconstructing objects and extracting high-quality surfaces play a vital role in the real world. Current 4D representations show the ability to render high-quality novel views for dynamic objects, but cannot reconstruct high-quality meshes due to their implicit or geometrically inaccurate representations. In this paper, we propose a novel representation that can reconstruct accurate meshes from sparse image input, named Dynamic 2D Gaussians (D-2DGS). We adopt 2D Gaussians for basic geometry representation and use sparse-controlled points to capture the 2D Gaussian's deformation. By extracting the object mask from the rendered high-quality image and masking the rendered depth map, we remove floaters that are prone to occur during reconstruction and can extract high-quality dynamic mesh sequences of dynamic objects. Experiments demonstrate that our D-2DGS is outstanding in reconstructing detailed and smooth high-quality meshes from sparse inputs. The code is available at https://github.com/hustvl/Dynamic-2DGS.