Prior-Enhanced Gaussian Splatting for Dynamic Scene Reconstruction from Casual Video

TL;DR

This paper presents an automatic method for reconstructing dynamic scenes from casual monocular videos by enhancing priors in Gaussian Splatting, leading to improved accuracy and rendering quality.

Contribution

It introduces a novel pipeline that leverages priors, segmentation, and tracking to enhance dynamic scene reconstruction without designing new scene representations.

Findings

Outperforms previous monocular dynamic scene reconstruction methods.

Produces more accurate and visually superior renderings.

Effectively captures fine geometry and coherent motion.

Abstract

We introduce a fully automatic pipeline for dynamic scene reconstruction from casually captured monocular RGB videos. Rather than designing a new scene representation, we enhance the priors that drive Dynamic Gaussian Splatting. Video segmentation combined with epipolar-error maps yields object-level masks that closely follow thin structures; these masks (i) guide an object-depth loss that sharpens the consistent video depth, and (ii) support skeleton-based sampling plus mask-guided re-identification to produce reliable, comprehensive 2-D tracks. Two additional objectives embed the refined priors in the reconstruction stage: a virtual-view depth loss removes floaters, and a scaffold-projection loss ties motion nodes to the tracks, preserving fine geometry and coherent motion. The resulting system surpasses previous monocular dynamic scene reconstruction methods and delivers visibly superior renderings