metadata
task_categories:
- image-to-video
SeeU45 Dataset
This repository contains the SeeU45 dataset introduced in the paper SeeU: Seeing the Unseen World via 4D Dynamics-aware Generation.
SeeU is a framework that learns continuous 4D dynamics from sparse and monocular 2D frames, enabling the generation of unseen visual content through a 2D $\to$ 4D $\to$ 2D learning process.
- Project Page: https://yuyuanspace.com/SeeU/
- GitHub Repository: https://github.com/pandayuanyu/SeeU
Sample Usage
Download Pre-Processed Dataset
You can download the pre-processed SeeU45 Dataset using the huggingface_hub CLI as shown in the official documentation:
pip install "huggingface_hub[hf_transfer]"
hf download pandaphd/SeeU45_PreProcessed --repo-type dataset --local-dir preproc/SeeU45
Citation
@article{Yuan_2025_SeeU,
title={{SeeU}: Seeing the Unseen World via 4D Dynamics-aware Generation},
author={Yuan, Yu and Wickremasinghe, Tharindu and Nadir, Zeeshan and Wang, Xijun and Chi, Yiheng and Chan, Stanley H.},
journal={arXiv preprint arXiv: 2512.03350},
year={2025}
}