Complet4R: Geometric Complete 4D Reconstruction

2026-03-28

AI Summary

Existing methods struggle to simultaneously ensure temporal coherence and geometric completeness in four-dimensional (4D) reconstruction of dynamic scenes under occlusion. This work proposes the first end-to-end unified framework that jointly models 4D reconstruction and geometric completion, departing from conventional pairwise reconstruction or local motion estimation strategies. Instead, it globally aggregates temporal context directly from video sequences using a decoder-only Transformer architecture. The model explicitly captures long-range temporal dependencies and generates, frame by frame, complete 3D geometry that includes occluded regions. Evaluated on a newly introduced benchmark for geometrically complete 4D reconstruction and a 3D point tracking task, the method achieves state-of-the-art performance.

Abstract

We introduce Complet4R, a novel end-to-end framework for Geometric Complete 4D Reconstruction, which aims to recover temporally coherent and geometrically complete reconstruction for dynamic scenes. Our method formalizes the task of Geometric Complete 4D Reconstruction as a unified framework of reconstruction and completion, by directly accumulating full contexts onto each frame. Unlike previous approaches that rely on pairwise reconstruction or local motion estimation, Complet4R utilizes a decoder-only transformer to operate all context globally directly from sequential video input, reconstructing a complete geometry for every single timestamp, including occluded regions visible in other frames. Our method demonstrates the state-of-the-art performance on our proposed benchmark for Geometric Complete 4D Reconstruction and the 3D Point Tracking task. Code will be released to support future research.

Problem

• Geometric Complete 4D Reconstruction
• dynamic scenes
• temporal coherence
• occlusion completion
• 4D reconstruction

Innovation

• Geometric Complete 4D Reconstruction
• decoder-only transformer
• temporal coherence
• global context accumulation
• dynamic scene reconstruction