|
|
--- |
|
|
language: |
|
|
- en |
|
|
license: mit |
|
|
multimodality: |
|
|
- video |
|
|
paperswithcode_id: null |
|
|
pretty_name: so100 Teleop Dataset |
|
|
tags: |
|
|
- robotics |
|
|
- imitation-learning |
|
|
- teleoperation |
|
|
- video |
|
|
task_categories: |
|
|
- robotics |
|
|
task_ids: |
|
|
- robotics-manipulation |
|
|
--- |
|
|
|
|
|
# Dataset Card for so100 Teleop Dataset |
|
|
|
|
|
## Dataset Description |
|
|
|
|
|
- **Repository:** [Add repository URL] |
|
|
- **Paper:** [Add paper reference if applicable] |
|
|
- **Point of Contact:** [Add contact information] |
|
|
|
|
|
### Dataset Summary |
|
|
|
|
|
This dataset contains teleoperation data for so100 robot manipulation tasks. The dataset includes: |
|
|
|
|
|
- **Videos:** RGB video recordings of robot manipulation |
|
|
- **Joint States:** Per-frame joint angle data |
|
|
- **Actions:** Robot joint actions derived from state differences |
|
|
- **Metadata:** Episode and frame indexing information |
|
|
|
|
|
### Supported Tasks and Leaderboards |
|
|
|
|
|
This dataset is designed for robot imitation learning and manipulation tasks. |
|
|
|
|
|
### Languages |
|
|
|
|
|
The dataset contains English language metadata. |
|
|
|
|
|
## Dataset Structure |
|
|
|
|
|
### Data Instances |
|
|
|
|
|
Each episode contains: |
|
|
- Video frames at 24.0 FPS |
|
|
- Joint state data for 6 joints |
|
|
- Action data for robot control |
|
|
- Timestamp information |
|
|
- Episode and frame indexing |
|
|
|
|
|
### Data Fields |
|
|
|
|
|
- `observation.state`: Joint angle data (float32, shape: [6]) |
|
|
- `action`: Robot joint actions (float32, shape: [6]) |
|
|
- `observation.images.main`: RGB video data (video, shape: [704, 1280, 3]) |
|
|
- `frame_index`: Frame index within episode (int64) |
|
|
- `episode_index`: Episode identifier (int64) |
|
|
- `index`: Global frame index (int64) |
|
|
- `task_index`: Task identifier (int64) |
|
|
- `timestamp`: Time from episode start (float32) |
|
|
- `next.done`: Episode termination flag (bool) |
|
|
|
|
|
### Data Splits |
|
|
|
|
|
- **Train:** 348 frames across 1 episode |
|
|
|
|
|
## Dataset Creation |
|
|
|
|
|
### Source Data |
|
|
|
|
|
#### Initial Data Collection and Normalization |
|
|
|
|
|
The dataset was created from teleoperation recordings of so100 robot manipulation tasks. |
|
|
|
|
|
#### Who are the source language producers? |
|
|
|
|
|
[Add information about data collection process] |
|
|
|
|
|
### Annotations |
|
|
|
|
|
#### Annotation process |
|
|
|
|
|
[Add information about annotation process if applicable] |
|
|
|
|
|
#### Who are the annotators? |
|
|
|
|
|
[Add information about annotators if applicable] |
|
|
|
|
|
### Personal and Sensitive Information |
|
|
|
|
|
[Add information about personal/sensitive data handling] |
|
|
|
|
|
## Additional Information |
|
|
|
|
|
### Dataset Curators |
|
|
|
|
|
[Add curator information] |
|
|
|
|
|
### Licensing Information |
|
|
|
|
|
This dataset is licensed under the MIT License. |
|
|
|
|
|
### Citation Information |
|
|
|
|
|
```bibtex |
|
|
@dataset{so100_teleop_dataset, |
|
|
title = {so100 Teleop Dataset}, |
|
|
author = {[Add author information]}, |
|
|
year = {2024}, |
|
|
url = {[Add dataset URL]} |
|
|
} |
|
|
``` |
|
|
|
|
|
### Contributions |
|
|
|
|
|
[Add contribution information] |
|
|
|
|
|
### Contact |
|
|
|
|
|
[Add contact information] |
|
|
|
