| 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](https://huggingface.co/papers/2512.03350). | |
| 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/](https://yuyuanspace.com/SeeU/) | |
| - **GitHub Repository:** [https://github.com/pandayuanyu/SeeU](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: | |
| ```bash | |
| pip install "huggingface_hub[hf_transfer]" | |
| hf download pandaphd/SeeU45_PreProcessed --repo-type dataset --local-dir preproc/SeeU45 | |
| ``` | |
| ## Citation | |
| ```bibtex | |
| @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} | |
| } | |
| ``` |