4DNeX-10M Dataset

📄 Paper  |  🚀 Project Page  |  💻 GitHub

Introduction

4DNeX-10M is a large-scale hybrid dataset introduced in the paper "4DNeX: Feed-Forward 4D Generative Modeling Made Easy". The dataset aggregates monocular videos from diverse sources, including both static and dynamic scenes, accompanied by high-quality pseudo 4D annotations generated using state-of-the-art 3D and 4D reconstruction methods. The dataset enables joint modeling of RGB appearance and XYZ geometry sequences as a unified 6D video representation, facilitating efficient and generalizable 4D scene generation.

Dataset Structure

4DNeX-10M/
├── dynamic_1/          # Dynamic videos from Pexels, annotated with MonST3R
├── dynamic_2/          # Dynamic videos from Video Generation Models (VDM), annotated with MonST3R
├── dynamic_3/          # Dynamic videos from Vimeo, annotated with MegaSaM
├── static_1/           # Static videos from DL3DV, annotated with DUSt3R
├── static_2/           # Static videos from RealEstate10K, annotated with DUSt3R
└── raw/                # Raw source videos from all the above sources with captions

Citation

If you use this dataset in your work, please cite our paper:

@article{chen20254dnex,
  title={4DNeX: Feed-Forward 4D Generative Modeling Made Easy},
  author={Chen, Zhaoxi and Liu, Tianqi and Zhuo, Long and Ren, Jiawei and Tao, Zeng and Zhu, He and Hong, Fangzhou and Pan, Liang and Liu, Ziwei},
  journal={arXiv preprint arXiv:2508.13154},
  year={2025}
}