ViDAR: Video Diffusion-Aware 4D Reconstruction From Monocular Inputs

TL;DR

ViDAR introduces a novel 4D reconstruction framework using diffusion models to generate multi-view supervision from monocular videos, significantly improving dynamic scene reconstruction quality.

Contribution

The paper presents ViDAR, a new method that leverages personalized diffusion models and a diffusion-aware loss for monocular 4D scene reconstruction.

Findings

Outperforms state-of-the-art methods on DyCheck benchmark.

Achieves better visual quality and geometric consistency.

Shows strong improvements in dynamic regions.

Abstract

Dynamic Novel View Synthesis aims to generate photorealistic views of moving subjects from arbitrary viewpoints. This task is particularly challenging when relying on monocular video, where disentangling structure from motion is ill-posed and supervision is scarce. We introduce Video Diffusion-Aware Reconstruction (ViDAR), a novel 4D reconstruction framework that leverages personalised diffusion models to synthesise a pseudo multi-view supervision signal for training a Gaussian splatting representation. By conditioning on scene-specific features, ViDAR recovers fine-grained appearance details while mitigating artefacts introduced by monocular ambiguity. To address the spatio-temporal inconsistency of diffusion-based supervision, we propose a diffusion-aware loss function and a camera pose optimisation strategy that aligns synthetic views with the underlying scene geometry. Experiments on DyCheck, a challenging benchmark with extreme viewpoint variation, show that ViDAR outperforms all state-of-the-art baselines in visual quality and geometric consistency. We further highlight ViDAR's strong improvement over baselines on dynamic regions and provide a new benchmark to compare performance in reconstructing motion-rich parts of the scene. Project page: https://vidar-4d.github.io