---
license: mit
tags:
- motion
- opticalflow
pretty_name: MateiralMotion
size_categories:
- 100B<n<1T
---


# Motion Estimation Datasets

This repository contains multiple datasets for motion estimation used in the paper  
**"Machine Learning Modeling for Multi-order Human Visual Motion Perception"**.

## Citation

For use, please cite:

```bibtex
@misc{sun2025machinelearningmodelingmultiorder,
  title={Machine Learning Modeling for Multi-order Human Visual Motion Processing},
  author={Zitang Sun and Yen-Ju Chen and Yung-Hao Yang and Yuan Li and Shin'ya Nishida},
  year={2025},
  eprint={2501.12810},
  archivePrefix={arXiv},
  primaryClass={cs.CV},
  url={https://arxiv.org/abs/2501.12810},
}
```

## Dataset Sources

- **Human Data:** Our human data are based on the Sintel-Slow and KITTI 2015 datasets.  
- **Motion Datasets:** The non-diffuse and diffuse material-based motion datasets are generated using the Kubric pipeline.
  
- [Kubric](https://github.com/google-research/kubric)
- [KITTI](https://www.cvlibs.net/datasets/kitti/)
- [MPI-Sintel](http://sintel.is.tue.mpg.de/)
- [Sintel-slow](https://www.cvlibs.net/projects/slow_flow/)
- [DAVIS](https://davischallenge.org/)


**Note:** You need to follow the respective licenses and policies for using the above datasets.

---