ry-5/data_grapes

---
license: apache-2.0
task_categories:
- robotics
tags:
- LeRobot
configs:
- config_name: default
data_files:
- split: train
path: data/chunk-000/episode_*.parquet
dataset_info:
features:
- name: timestamp
dtype: float64
- name: frame_index
dtype: int64
- name: episode_index
dtype: int64
- name: index
dtype: int64
- name: task_index
dtype: int64
- name: observation.state
sequence:
dtype: float64
length: 7
- name: observation.velocity
sequence:
dtype: float64
length: 7
- name: observation.effort
sequence:
dtype: float64
length: 7
- name: observation.weight
sequence:
dtype: float64
length: 1
- name: action
sequence:
dtype: float64
length: 7
splits:
- name: train
num_bytes: 16603644
num_examples: 120000
download_size: 16603644
dataset_size: 16603644
---

# FR5 Garlic Manipulation Dataset

This dataset contains demonstrations collected on an FR5 robot teleoperated to manipulate garlic. Each episode pairs synchronized joint trajectories (`observation.state`, `action`) with base and wrist camera videos. The layout follows the LeRobot dataset convention with chunked parquet files and per-camera mp4 streams.

## Dataset Details

- **Total Episodes**: 100
- **Total Frames**: 120,000
- **FPS**: ~30.01 Hz
- **Episode Length**: ~1200 frames per episode
- **Task**: Garlic pick and place manipulation

## Data Structure

Each episode contains:
- **Joint States** (`observation.state`): 7D joint positions (6 arm joints + 1 gripper) in radians
- **Actions** (`action`): 7D target joint positions (6 arm joints + 1 gripper) in radians
- **Additional Observations**: velocity, effort, and weight measurements
- **Videos**: Synchronized base and wrist camera videos (MP4 format)
- **Metadata**: Timestamps, frame indices, episode indices, and task indices

## File Structure

```
demos_garlic_mod/
├── data/
│ └── chunk-000/
│ └── episode_*.parquet
├── videos/
│ └── chunk-000/
│ ├── observation.images.cam_base/
│ │ └── episode_*.mp4
│ ├── observation.images.cam_wrist/
│ │ └── episode_*.mp4
└── meta/
├── info.json
├── episodes.jsonl
├── episodes_stats.jsonl
└── tasks.jsonl
```

## Usage

This dataset is compatible with the LeRobot framework and can be loaded using standard LeRobot data loaders for training imitation learning models.