# ForceBand EMG2Force: Wrist Surface EMG and IMU with Per-Finger Fingertip Force and RGB Video (He, Malik et al.)

## Facts

- Provider: Botao He, Zhi Wang, Linna Kuang, Ishaan Ghosh, Jitendra Malik, Cornelia Fermuller et al.
- License: Apache-2.0 (https://huggingface.co/datasets/boootaooo/forceband-emg2force)
- Size: 253 files
- Format: CSV, JSON, MP4 (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/boootaooo/forceband-emg2force
- Access: public, readable without an account
- Category: Physiology
- Source: https://arxiv.org/abs/2606.26093
- Page: https://scigantic.com/archive/d364f67f-6d74-5a96-8bd3-17efcdbcf7e8
- Keywords: Physiology, EMG, CSV, JSON, MP4, Hugging Face

## About

The pretraining data of the EMG2Force model from ForceBand: Learning Forceful Manipulation with sEMG (He, Wang, Kuang, Ghosh, Malik, Fermuller and colleagues, arXiv:2606.26093). A wrist band records 8-channel surface electromyography (sEMG) and a 10-channel inertial measurement unit (IMU: orientation quaternion, gyroscope and accelerometer) together with fingertip force for each of the five fingers, all at 1 kHz; RGB video from one or more cameras is synchronized by per-frame timestamps. The release has 53 recordings from October 2025 to April 2026, 8.61 hours of cleaned EMG, IMU and force (8.39 hours valid), 6.37 hours of it with video, about 7.7 GB. Each segment folder holds timeseries.hdf5 (timestamps, emg, imu, force, validity masks, trial and task indices) plus H.264 MP4 video and frame timestamps; manifest.csv, recording_log.csv and summary.json index the segments. Load with h5py and huggingface_hub. Uses: regressing grip and finger force from forearm muscle activity, EMG-based hand interfaces, and teaching robots forceful manipulation from human demonstrations. Video shows participants' faces. The authors state the dataset is released under Apache-2.0, the same as the ForceBand code (LICENSE file in the repository). Cite He et al., ForceBand, 2026.

## Load the data

```python
import pandas as pd
df = pd.read_csv("hf://datasets/boootaooo/forceband-emg2force/manifest.csv")
df.head()
```

## Data structure

File types in sampled files: CSV, JSON, MP4

Example file: `manifest.csv`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/d364f67f-6d74-5a96-8bd3-17efcdbcf7e8
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/d364f67f-6d74-5a96-8bd3-17efcdbcf7e8
- Over MCP: `get_schema_card` with id `d364f67f-6d74-5a96-8bd3-17efcdbcf7e8` at https://api.scigantic.com/mcp

## Related topics

- [JSON datasets](https://scigantic.com/topics/json)
- [CSV datasets](https://scigantic.com/topics/csv)
- [Physiology datasets](https://scigantic.com/topics/physiology)
- [EMG datasets](https://scigantic.com/topics/emg)
- [MP4 datasets](https://scigantic.com/topics/mp4)
