Rynn乐云具身_抓取任务

# Grasp Strawberry Into Basket Dataset

## 项目简介

本数据集用于机器人抓取草莓并放入杯子/篮子的任务学习。数据集包含多个采集样本，每个样本记录了完整的抓取操作过程。

## 任务描述

**任务目标**：Pick up the strawberries and put them in the cup（抓取草莓并放入杯中）

## 数据集结构

数据集包含 **108个样本**，每个样本文件夹包含以下文件：

```
├── N8rXXXXXX/
│ ├── metadata.json # 元数据信息
│ ├── observation.images.front.mp4 # 前置摄像头视频
│ ├── observation.images.wrist.mp4 # 手腕摄像头视频
│ └── timeseries.parquet # 时序数据（机械臂状态、动作等）
```

## 数据格式说明

### metadata.json
包含样本的元信息：
- `task_description`: 任务描述
- `task_prompt`: 任务提示
- `total_frames`: 总帧数
- `fps`: 帧率（30fps）
- `collection_time`: 采集时间

### 视频数据
- **front视频**: 固定前置摄像头拍摄的操作场景
- **wrist视频**: 机械臂手腕摄像头拍摄的抓取视角

### timeseries.parquet
包含每一帧的时序数据，如机械臂关节位置、末端执行器状态等。

## 使用场景

- 机器人模仿学习（Imitation Learning）
- 视觉-动作策略学习
- 机器人操作技能训练

## 许可证

请遵循相关数据使用规范。

# Grasp Strawberry Into Basket Dataset

## Project Overview
This dataset is designed for robotic learning tasks involving grasping strawberries and placing them into cups or baskets. It includes multiple collected samples, each recording the complete grasping operation process.

## Task Description
**Task Objective**: Grasp strawberries and place them into cups or baskets

## Dataset Structure
This dataset contains **108 samples**, with each sample folder including the following files:

├── N8rXXXXXX/
│ ├── metadata.json # Metadata information
│ ├── observation.images.front.mp4 # Front-facing camera video
│ ├── observation.images.wrist.mp4 # Wrist-mounted camera video
│ └── timeseries.parquet # Time-series data (robot arm states, actions, etc.)

## Data Format Specifications

### metadata.json
Contains sample metadata:
- `task_description`: Task description
- `task_prompt`: Task prompt
- `total_frames`: Total number of frames
- `fps`: Frame rate (30fps)
- `collection_time`: Collection time

### Video Data
- **Front video**: Operational scene captured by a fixed front-facing camera
- **Wrist video**: Grasping perspective captured by the camera mounted on the robot's wrist

### timeseries.parquet
Contains time-series data for each frame, including robot arm joint positions, end-effector states, and other relevant information.

## Application Scenarios
- Robot Imitation Learning
- Visual-Action Policy Learning
- Robotic Manipulation Skill Training

## License
Please comply with relevant data usage specifications.

搜集汇总

背景与挑战

背景概述

该数据集专为机器人学习抓取草莓并放入杯中的任务设计，包含108个样本，每个样本提供元数据、双视角视频及时序数据。它适用于机器人模仿学习、视觉动作策略学习和操作技能训练等领域，遵循Apache License 2.0许可。