Summary
Syn4D is a multiview synthetic dataset of dynamic scenes introduced to address the scarcity of high-quality ground-truth data for dense 3D reconstruction and tracking from monocular video. The dataset provides accurate camera motion, depth maps, dense point tracking, and parametric human pose annotations. Its defining feature is the ability to unproject any pixel into full 3D at any timestamp and from any camera viewpoint, enabling dense, complete geometric supervision that real-world captures cannot offer. The authors, including researchers from the University of Oxford and Naver Labs Europe, evaluate the dataset on multiple downstream tasks: 4D scene reconstruction, 3D point tracking, geometry-aware camera retargeting, and human pose estimation. Experimental results demonstrate Syn4D's utility as a benchmark and training resource for advancing dynamic scene understanding and spatiotemporal modeling research. The paper is available on arXiv as 2605.05207.
Paper Overview
- Field: Computer Vision
- Authors: Zeren Jiang, Yushi Lan, Yihang Luo, Yufan Deng, Zihang Lai, Edgar Sucar, Christian Rupprecht, Iro Laina, Diane Larlus, Chuanxia Zheng, Andrea Vedaldi
- Published: 2026-05-06
- arXiv: 2605.05207
Abstract
Dense 3D reconstruction and tracking of dynamic scenes from monocular video remains an important open challenge in computer vision. Progress in this area has been constrained by the scarcity of high-quality datasets with dense, complete, and accurate geometric annotations. To address this limitation, the authors introduce Syn4D, a multiview synthetic dataset of dynamic scenes that includes ground-truth camera motion, depth maps, dense tracking, and parametric human pose annotations.
A key feature of Syn4D is the ability to unproject any pixel into 3D to any time and to any camera. The authors conduct extensive evaluations across multiple downstream tasks to demonstrate the utility and effectiveness of the proposed dataset, including:
- 4D scene reconstruction
- 3D point tracking
- Geometry-aware camera retargeting
- Human pose estimation
The experimental results highlight Syn4D's potential to facilitate research in dynamic scene understanding and spatiotemporal modeling.
---
*Auto-collected on 2026-05-08 from zhichai.net.*
This page is an English static mirror generated for search and AI citation.
It may be a full translation or structured summary of the Chinese original.
Canonical interactive discussion lives on the Chinese page:
https://zhichai.net/topic/177619578