---
language:
- en
license: cc-by-4.0
size_categories:
- 100K<n<1M
task_categories:
- image-to-image
pretty_name: X2C
tags:
- image
- datasets
- croissant
- pandas
- json
---

---

# Dataset Card for X2C

<!-- Provide a quick summary of the dataset. -->

X2C (Anything to Control) is a dataset featuring nuanced robotic facial expressions for realistic humanoid imitation and facial expression generation. It contains 100,000 images of a robot displaying diverse facial expressions, 
each annotated with 30 control values that can be applied to the physical robot to recreate the corresponding expression.

## Dataset Details

- For more details, please refer to our paper: [X2C: A Dataset Featuring Nuanced Facial Expressions for Realistic Humanoid Imitation](https://huggingface.co/papers/2505.11146)
- Source code is available at: [X2CNet](https://github.com/lipzh5/X2CNet)

### Dataset Description

<!-- Provide a longer summary of what this dataset is. -->


- **Curated by:** This project was primarily conducted and received ethics approval via Macquarie University. 
- **Language(s):** English
- **License:**  Creative Commons Attribution 4.0 (CC BY 4.0)



## Citation [optional]

<!-- If there is a paper or blog post introducing the dataset, the APA and Bibtex information for that should go in this section. -->

**BibTeX:**

@article{li2025x2c,
  title={X2C: A Dataset Featuring Nuanced Facial Expressions for Realistic Humanoid Imitation},
  author={Li, Peizhen and Cao, Longbing and Wu, Xiao-Ming and Yang, Runze and Yu, Xiaohan},
  journal={arXiv preprint arXiv:2505.11146},
  year={2025}
}


## Dataset Card Authors 

peizhen.li1@hdr.mq.edu.au



## Dataset Card Contact

For any comments or questions, please email [Peizhen Li](mailto:peizhen.li1@hdr.mq.edu.au)