Update README.md
Browse files
README.md
CHANGED
|
@@ -5,33 +5,25 @@ dataset_info:
|
|
| 5 |
- name: image
|
| 6 |
dtype: image
|
| 7 |
- name: actuated_angle
|
| 8 |
-
|
| 9 |
-
- name: '0'
|
| 10 |
-
dtype: int32
|
| 11 |
-
- name: '1'
|
| 12 |
-
dtype: int32
|
| 13 |
splits:
|
| 14 |
- name: train
|
| 15 |
-
num_bytes:
|
| 16 |
num_examples: 5
|
| 17 |
-
download_size:
|
| 18 |
-
dataset_size:
|
| 19 |
- config_name: small
|
| 20 |
features:
|
| 21 |
- name: image
|
| 22 |
dtype: image
|
| 23 |
- name: actuated_angle
|
| 24 |
-
|
| 25 |
-
- name: '0'
|
| 26 |
-
dtype: int32
|
| 27 |
-
- name: '1'
|
| 28 |
-
dtype: int32
|
| 29 |
splits:
|
| 30 |
- name: train
|
| 31 |
-
num_bytes:
|
| 32 |
num_examples: 2
|
| 33 |
-
download_size:
|
| 34 |
-
dataset_size:
|
| 35 |
configs:
|
| 36 |
- config_name: full
|
| 37 |
data_files:
|
|
@@ -42,4 +34,120 @@ configs:
|
|
| 42 |
- split: train
|
| 43 |
path: small/train-*
|
| 44 |
default: true
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 45 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 5 |
- name: image
|
| 6 |
dtype: image
|
| 7 |
- name: actuated_angle
|
| 8 |
+
sequence: int32
|
|
|
|
|
|
|
|
|
|
|
|
|
| 9 |
splits:
|
| 10 |
- name: train
|
| 11 |
+
num_bytes: 5561735
|
| 12 |
num_examples: 5
|
| 13 |
+
download_size: 5564250
|
| 14 |
+
dataset_size: 5561735
|
| 15 |
- config_name: small
|
| 16 |
features:
|
| 17 |
- name: image
|
| 18 |
dtype: image
|
| 19 |
- name: actuated_angle
|
| 20 |
+
sequence: int32
|
|
|
|
|
|
|
|
|
|
|
|
|
| 21 |
splits:
|
| 22 |
- name: train
|
| 23 |
+
num_bytes: 2219196
|
| 24 |
num_examples: 2
|
| 25 |
+
download_size: 2221457
|
| 26 |
+
dataset_size: 2219196
|
| 27 |
configs:
|
| 28 |
- config_name: full
|
| 29 |
data_files:
|
|
|
|
| 34 |
- split: train
|
| 35 |
path: small/train-*
|
| 36 |
default: true
|
| 37 |
+
license: mit
|
| 38 |
+
task_categories:
|
| 39 |
+
- image-feature-extraction
|
| 40 |
+
tags:
|
| 41 |
+
- robotics
|
| 42 |
+
size_categories:
|
| 43 |
+
- 1M<n<10M
|
| 44 |
---
|
| 45 |
+
|
| 46 |
+
|
| 47 |
+
# 🧠 Open Humnoid Actuated Face Dataset
|
| 48 |
+
|
| 49 |
+
<img src="https://huggingface.co/datasets/iamirulofficial/Test2/resolve/main/imgesFace.png" alt="Sample Face Image" width="400"/>
|
| 50 |
+
|
| 51 |
+
## Dataset Summary
|
| 52 |
+
|
| 53 |
+
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.
|
| 54 |
+
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.
|
| 55 |
+
|
| 56 |
+
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.
|
| 57 |
+
|
| 58 |
+
---
|
| 59 |
+
|
| 60 |
+
## Dataset Structure
|
| 61 |
+
|
| 62 |
+
Each data sample contains:
|
| 63 |
+
|
| 64 |
+
- **Actuated Angles**:
|
| 65 |
+
- 16 actuator values representing positions/angles controlling the face and eyes.
|
| 66 |
+
|
| 67 |
+
- **Face Image**:
|
| 68 |
+
- RGB image of the humanoid face (resolution: **[FILL_RESOLUTION]**) matching the actuator positions.
|
| 69 |
+
|
| 70 |
+
- **Actuator Position Images**:
|
| 71 |
+
- Annotated actuator mapping images:
|
| 72 |
+
- Full Face Actuator Map.
|
| 73 |
+
- Eye Actuator Map.
|
| 74 |
+
|
| 75 |
+
### Actuator Details
|
| 76 |
+
|
| 77 |
+
| Index | Actuator | Index | Actuator |
|
| 78 |
+
| :---: | ------------------------ | :---: | -------------------- |
|
| 79 |
+
| 00 | Cheek - Left | 08 | Eyelid Upper - Right |
|
| 80 |
+
| 01 | Cheek - Right | 09 | Eyelid Lower - Right |
|
| 81 |
+
| 02 | Eyeball Sideways - Left | 10 | Forehead - Right |
|
| 82 |
+
| 03 | Eyeball Up Down - Left | 11 | Forehead - Left |
|
| 83 |
+
| 04 | Eyelid Upper - Left | 12 | Upper Nose |
|
| 84 |
+
| 05 | Eyelid Lower - Left | 13 | Eyebrow - Right |
|
| 85 |
+
| 06 | Eyeball Up Down - Right | 14 | Jaw |
|
| 86 |
+
| 07 | Eyeball Sideways - Right | 15 | Eyebrow - Left |
|
| 87 |
+
|
| 88 |
+
---
|
| 89 |
+
|
| 90 |
+
|
| 91 |
+
**Actuator Mapping Images**:
|
| 92 |
+
|
| 93 |
+
**Full Face Actuator Map**
|
| 94 |
+
|
| 95 |
+
<p align="center"> <img src="https://huggingface.co/datasets/iamirulofficial/Test2/resolve/main/Screenshot%202025-05-02%20at%204.04.28%E2%80%AFPM.png" alt="Full Face Actuator Map" width="35%"/> <img src="https://huggingface.co/datasets/iamirulofficial/Test2/resolve/main/Screenshot%202025-05-02%20at%204.03.56%E2%80%AFPM.png" alt="Eye Actuator Map" width="35%"/> </p>
|
| 96 |
+
|
| 97 |
+
---
|
| 98 |
+
|
| 99 |
+
## Dataset Size
|
| 100 |
+
|
| 101 |
+
- **Total Samples**: [FILL_NUMBER_OF_SAMPLES]
|
| 102 |
+
- **Splits**: No predefined splits. However, users can load either the full dataset or a smaller subset using the provided configuration.
|
| 103 |
+
|
| 104 |
+
---
|
| 105 |
+
|
| 106 |
+
## Usage
|
| 107 |
+
|
| 108 |
+
You can load either the full dataset or a small subset for quick experiments:
|
| 109 |
+
|
| 110 |
+
```python
|
| 111 |
+
from datasets import load_dataset
|
| 112 |
+
|
| 113 |
+
# Load the small config (with actuated angles)
|
| 114 |
+
ds = load_dataset("./", name="small", data_dir=".") # Ideally use repo ID
|
| 115 |
+
|
| 116 |
+
# push everything—including arrow files—to your hub repo
|
| 117 |
+
ds.push_to_hub("iamirulofficial/TestImageDataset")
|
| 118 |
+
````
|
| 119 |
+
|
| 120 |
+
---
|
| 121 |
+
|
| 122 |
+
## Data Collection Method
|
| 123 |
+
|
| 124 |
+
Stay tuned — more details will be available in the upcoming paper describing the RL training process, actuator control policies, and data generation methodology.
|
| 125 |
+
|
| 126 |
+
---
|
| 127 |
+
|
| 128 |
+
## License
|
| 129 |
+
|
| 130 |
+
This dataset is released under the **MIT License**.
|
| 131 |
+
|
| 132 |
+
---
|
| 133 |
+
|
| 134 |
+
## Citation
|
| 135 |
+
|
| 136 |
+
If you use this dataset, please cite:
|
| 137 |
+
|
| 138 |
+
```bibtex
|
| 139 |
+
@misc{faceheadactuated2025,
|
| 140 |
+
title = {Open Humnoid Actuated Face Dataset},
|
| 141 |
+
author = {Your Name or Organization},
|
| 142 |
+
year = {2025},
|
| 143 |
+
url = {https://huggingface.co/datasets/YOUR_DATASET_PATH}
|
| 144 |
+
}
|
| 145 |
+
```
|
| 146 |
+
|
| 147 |
+
---
|
| 148 |
+
|
| 149 |
+
## Acknowledgements
|
| 150 |
+
|
| 151 |
+
Special thanks to the team developing the **i2Head InMoov** platform and contributors to the silicon skin design.
|
| 152 |
+
|
| 153 |
+
|