---
license: cc-by-nc-4.0
tags:
- model_hub_mixin
- pytorch_model_hub_mixin
---

# MoSE3: Learning World-Space SE(3) at Every Pixel

[Paper](https://arxiv.org/abs/2610.03716) | [Project page](https://mose3-tracker.github.io/) | [Code](https://github.com/mose3-tracker/MoSE3)

**MoSE3** is a feed-forward model that predicts world-space **SE(3) motion for every pixel** of a
monocular RGB video, generalizing across **rigid, articulated and deformable objects**.

## Usage

```python
from mose3.models.mose3 import MoSE3

model = MoSE3.from_pretrained("mose3-tracker/MoSE3", strict=True).cuda().eval()
```

The [code repository](https://github.com/mose3-tracker/MoSE3) has the full inference and visualization scripts.

## License

The weights are released under [CC BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/)
(non-commercial). They include the weights of π³, which are released under the same license.

## Citation

```bibtex
@inproceedings{cheng2026mose3,
  title     = {{MoSE3}: Learning World-Space {SE(3)} at Every Pixel},
  author    = {Cheng, Jiahuan and Li, Zhiyi and Xia, Tian and
               Cai, Ruojin and Du, Yilun and Wang, Qianqian},
  booktitle = {Advances in Neural Information Processing Systems (NeurIPS)},
  year      = {2026}
}
```