---
license: bsd-3-clause
pipeline_tag: robotics
tags:
- fingernet
- metafinger
- multimodal
- onnx
- pytorch
library_name: transformers
datasets:
- asRobotics/fingernet-100k
---

<p align="center">
  <img src="/static-proxy?url=https%3A%2F%2Fcdn-uploads.huggingface.co%2Fproduction%2Fuploads%2F670336f25ce58dd0c33e32e3%2FXa0-nYJGWwQs78dAYqMoN.jpeg"
       style="width:100%; max-width:1000px; height:auto;">
</p>

# FingerNet

## Model Description

FingerNet is an MLP model designed for the asFinger. It can predict both 6D force and 3D shape (mesh nodes) from the 6D motion of the asFinger.

Try it out on the [Spaces Demo](https://huggingface.co/spaces/han-xudong/fingernet-demo)!

- Developer: Xudong Han, Ning Guo, Xiaobo Liu, Fang Wan, and Chaoyang Song.
- Model type: MLP
- License: BSD-3-Clause
- Github: [https://github.com/han-xudong/metafinger](https://github.com/han-xudong/metafinger)

## Intended Use

This model is intended for researchers and developers working in robotics and tactile sensing. It can be used to enhance the capabilities of asFingers in applications such as robotic manipulation, haptics, and human-robot interaction.

To load the model:

```python
# Example code to load safetensors
from transformers import AutoModel

model = AutoModel.from_pretrained("asRobotics/fingernet", trust_remote_code=True)
x = torch.zeros((1, 6))  # Example input: batch size of 1, 6D motion
output = model(x)
```

Or to load the ONNX version:

```python
# Example code to load onnx
import onnxruntime as ort
import numpy as np
from huggingface_hub import hf_hub_download

onnx_model_path = hf_hub_download(repo_id="asRobotics/fingernet", filename="model.onnx")
ort_session = ort.InferenceSession(onnx_model_path)
x = np.zeros((1, 6)).astype(np.float32)  # Example input: batch size of 1, 6D motion
outputs = ort_session.run(None, {"motion": x})
```

## Training Data

The model was trained on the [FingerNet-100K](https://huggingface.co/datasets/asRobotics/fingernet-100k), which includes a variety of motion, force, and shape data collected by finite element simulations.

## Citation

If you use this model in your research, please cite the following papers:

```bibtex
@article{liu2024proprioceptive,
  title={Proprioceptive learning with soft polyhedral networks},
  author={Liu, Xiaobo and Han, Xudong and Hong, Wei and Wan, Fang and Song, Chaoyang},
  journal={The International Journal of Robotics Research},
  volume = {43},
  number = {12},
  pages = {1916-1935},
  year = {2024},
  publisher={SAGE Publications Sage UK: London, England},
  doi = {10.1177/02783649241238765}
}
```

[](https://arxiv.org/abs/2308.08538)

```bibtex
@article{wu2025magiclaw,
  title={MagiClaw: A Dual-Use, Vision-Based Soft Gripper for Bridging the Human Demonstration to Robotic Deployment Gap},
  author={Wu, Tianyu and Han, Xudong and Sun, Haoran and Zhang, Zishang and Huang, Bangchao and Song, Chaoyang and Wan, Fang},
  journal={arXiv preprint arXiv:2509.19169},
  year={2025}
}
```