iamirulofficial/Test2

🧠 Open Humnoid Actuated Face Dataset

Dataset Summary

The Open Humnoid Actuated Face Dataset has been created to support researchers and developers working on face-head actuated simulations for robotics, reinforcement learning, and human-computer interaction.
The data was collected during a reinforcement learning (RL) training process where the agent’s goal was to replicate human facial expressions using a humanoid head model.

The dataset pairs actuator angle positions with corresponding images of a humanoid face (modified i2Head InMoov with a silicon skin). The system consists of 16 actuators controlling facial regions and eyes.

---

Dataset Structure

Each data sample contains:

• Actuated Angles:

  • 16 actuator values representing positions/angles controlling the face and eyes.

• Face Image:

  • RGB image of the humanoid face (resolution: [FILL_RESOLUTION]) matching the actuator positions.

• Actuator Position Images:

  • Annotated actuator mapping images:
    • Full Face Actuator Map.
    • Eye Actuator Map.

Actuator Details

Index	Actuator	Index	Actuator
00	Cheek - Left	08	Eyelid Upper - Right
01	Cheek - Right	09	Eyelid Lower - Right
02	Eyeball Sideways - Left	10	Forehead - Right
03	Eyeball Up Down - Left	11	Forehead - Left
04	Eyelid Upper - Left	12	Upper Nose
05	Eyelid Lower - Left	13	Eyebrow - Right
06	Eyeball Up Down - Right	14	Jaw
07	Eyeball Sideways - Right	15	Eyebrow - Left

---

Actuator Mapping Images:

Full Face Actuator Map

---

Dataset Size

• Total Samples: [FILL_NUMBER_OF_SAMPLES]
• Splits: No predefined splits. However, users can load either the full dataset or a smaller subset using the provided configuration.

---

Usage

You can load either the full dataset or a small subset for quick experiments:

from datasets import load_dataset

# Load the small config (with actuated angles)
ds = load_dataset("./", name="small", data_dir=".")  # Ideally use repo ID

# push everything—including arrow files—to your hub repo
ds.push_to_hub("iamirulofficial/TestImageDataset")

---

Data Collection Method

Stay tuned — more details will be available in the upcoming paper describing the RL training process, actuator control policies, and data generation methodology.

---

License

This dataset is released under the MIT License.

---

Citation

If you use this dataset, please cite:

@misc{faceheadactuated2025,
  title = {Open Humnoid Actuated Face Dataset},
  author = {Your Name or Organization},
  year = {2025},
  url = {https://huggingface.co/datasets/YOUR_DATASET_PATH}
}

---

Acknowledgements

Special thanks to the team developing the i2Head InMoov platform and contributors to the silicon skin design.