WorldTree: Towards 4D Dynamic Worlds from Monocular Video using Tree-Chains

2026-02-12

AI Summary

Existing methods for dynamic scene reconstruction from monocular videos lack a unified spatiotemporal decomposition framework, leading to inefficient coupling between temporal optimization and spatial modeling. To address this limitation, this work proposes WorldTree, a novel framework that introduces a Temporal Partitioning Tree (TPT) for coarse-to-fine hierarchical temporal decomposition and designs a Spatial Ancestry Chain (SAC) to recursively model spatial structural inheritance. By jointly learning complementary spatial dynamics and motion-specific representations, WorldTree achieves decoupled yet unified 4D reconstruction. Experimental results on the NVIDIA-LS and DyCheck benchmarks demonstrate consistent improvements over state-of-the-art methods, with LPIPS metrics enhanced by 8.26% and 9.09%, respectively.

Technology Category

• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Computer Vision: Motion & Tracking
• Planning, Routing, and Scheduling: Temporal Planning

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Dynamic reconstruction has achieved remarkable progress, but there remain challenges in monocular input for more practical applications. The prevailing works attempt to construct efficient motion representations, but lack a unified spatiotemporal decomposition framework, suffering from either holistic temporal optimization or coupled hierarchical spatial composition. To this end, we propose WorldTree, a unified framework comprising Temporal Partition Tree (TPT) that enables coarse-to-fine optimization based on the inheritance-based partition tree structure for hierarchical temporal decomposition, and Spatial Ancestral Chains (SAC) that recursively query ancestral hierarchical structure to provide complementary spatial dynamics while specializing motion representations across ancestral nodes. Experimental results on different datasets indicate that our proposed method achieves 8.26% improvement of LPIPS on NVIDIA-LS and 9.09% improvement of mLPIPS on DyCheck compared to the second-best method. Code: https://github.com/iCVTEAM/WorldTree.

Problem

• dynamic reconstruction
• monocular video
• spatiotemporal decomposition
• temporal optimization
• hierarchical spatial composition

Innovation

• Temporal Partition Tree
• Spatial Ancestral Chains
• 4D Dynamic Reconstruction
• Monocular Video
• Hierarchical Spatiotemporal Decomposition