English

Predicting 3D representations for Dynamic Scenes

Computer Vision and Pattern Recognition 2025-01-29 v1

Abstract

We present a novel framework for dynamic radiance field prediction given monocular video streams. Unlike previous methods that primarily focus on predicting future frames, our method goes a step further by generating explicit 3D representations of the dynamic scene. The framework builds on two core designs. First, we adopt an ego-centric unbounded triplane to explicitly represent the dynamic physical world. Second, we develop a 4D-aware transformer to aggregate features from monocular videos to update the triplane. Coupling these two designs enables us to train the proposed model with large-scale monocular videos in a self-supervised manner. Our model achieves top results in dynamic radiance field prediction on NVIDIA dynamic scenes, demonstrating its strong performance on 4D physical world modeling. Besides, our model shows a superior generalizability to unseen scenarios. Notably, we find that our approach emerges capabilities for geometry and semantic learning.

Keywords

Cite

@article{arxiv.2501.16617,
  title  = {Predicting 3D representations for Dynamic Scenes},
  author = {Di Qi and Tong Yang and Beining Wang and Xiangyu Zhang and Wenqiang Zhang},
  journal= {arXiv preprint arXiv:2501.16617},
  year   = {2025}
}
R2 v1 2026-06-28T21:21:05.924Z