V-DPM: 4D Video Reconstruction with Dynamic Point Maps

TL;DR

This paper introduces V-DPM, a novel dynamic point map framework for 4D video reconstruction that leverages video input and pretrained models to achieve state-of-the-art results in dynamic scene 3D and 4D reconstruction.

Contribution

The paper formulates DPMs for video input, enabling neural prediction and reuse of pretrained models, and demonstrates their effectiveness on top of VGGT for dynamic scene reconstruction.

Findings

Achieves state-of-the-art 3D and 4D reconstruction performance.

Recovers full 3D motion of scene points.

Adapts static scene reconstructor to dynamic scenes with synthetic data.

Abstract

Powerful 3D representations such as DUSt3R invariant point maps, which encode 3D shape and camera parameters, have significantly advanced feed forward 3D reconstruction. While point maps assume static scenes, Dynamic Point Maps (DPMs) extend this concept to dynamic 3D content by additionally representing scene motion. However, existing DPMs are limited to image pairs and, like DUSt3R, require post processing via optimization when more than two views are involved. We argue that DPMs are more useful when applied to videos and introduce V-DPM to demonstrate this. First, we show how to formulate DPMs for video input in a way that maximizes representational power, facilitates neural prediction, and enables reuse of pretrained models. Second, we implement these ideas on top of VGGT, a recent and powerful 3D reconstructor. Although VGGT was trained on static scenes, we show that a modest amount of synthetic data is sufficient to adapt it into an effective V-DPM predictor. Our approach achieves state of the art performance in 3D and 4D reconstruction for dynamic scenes. In particular, unlike recent dynamic extensions of VGGT such as P3, DPMs recover not only dynamic depth but also the full 3D motion of every point in the scene.