jnsungp/unitree-g1-robocasa-pick-apple-bowl-1k

--- license: cc-by-4.0 task_categories: - robotics tags: - unitree-g1 - pick-and-place - simulation - curobo - motion-planning size_categories: - 100K<n<1M language: - en pretty_name: Unitree G1 Apple Pick and Place Simulation Dataset --- # Unitree G1 Apple Pick and Place Simulation Dataset <div align="center"> <table> <tr> <td align="center" style="vertical-align: top; padding: 5px;"> <img src="https://huggingface.co/datasets/jnsungp/unitree-g1-robocasa-pick-apple-bowl-1k/resolve/main/frontview.png" width="500px" alt="Front View"/> <br><b>Front View (Global)</b> </td> <td align="center" style="vertical-align: top; padding: 5px;"> <img src="https://huggingface.co/datasets/jnsungp/unitree-g1-robocasa-pick-apple-bowl-1k/resolve/main/sideview.png" width="500px" alt="Side View"/> <br><b>Side View (Profile)</b> </td> </tr> <tr> <td align="center" style="vertical-align: top; padding: 5px;"> <img src="https://huggingface.co/datasets/jnsungp/unitree-g1-robocasa-pick-apple-bowl-1k/resolve/main/birdview.png" width="500px" alt="Top-Down View"/> <br><b>Top-Down View (Bird's Eye)</b> </td> <td align="center" style="vertical-align: top; padding: 5px;"> <img src="https://huggingface.co/datasets/jnsungp/unitree-g1-robocasa-pick-apple-bowl-1k/resolve/main/rs_view.png" width="500px" alt="Ego-Centric View"/> <br><b>Ego-Centric View (Robot POV)</b> </td> </tr> <tr> <td colspan="2" align="center" style="padding-top: 20px;"> <i>Multi-view perspectives of the Unitree G1 performing the pick-and-place task.</i> </td> </tr> </table> </div> ## Dataset Description The **Unitree G1 Apple Pick and Place Simulation Dataset** contains **957 high-quality trajectories** of a Unitree G1 humanoid robot performing pick-and-place tasks in simulation. The robot picks up a red apple from a table and places it into a bowl using bilateral arms and dexterous tri-finger hands. All trajectories are generated using CuRobo motion planning for collision-free, smooth execution. **Key Features:** - 957 successful simulation trajectories (100% success rate) - 28-DOF control: bilateral arms (7+7) + dexterous hands (7+7) - 256×256 RGB video at 20 FPS (ego view) - CuRobo motion planning (collision-free trajectories) - MuJoCo + RoboCasa simulation environment This dataset is ready for commercial and non-commercial use. ## Dataset Owner **Junsung Park** ([@jnsungp](https://huggingface.co/jnsungp)) Seoul National University [e-mail] ## License Creative Commons Attribution 4.0 International (CC BY 4.0) ## Dataset Format | Modality | Type | Shape | Description | |----------|------|-------|-------------| | **Observation** | `float32` | `(28,)` | Joint positions (radians) for arms + hands | | **Action** | `float32` | `(28,)` | Target joint positions | | **Video** | RGB | `(256, 256, 3)` | Ego view, 20 FPS, H.264 | | **Language** | `string` | - | _"Pick up the red apple and place it on the bowl"_ | ### Joint Configuration (28-DOF) | Body Part | DOF | Description | |-----------|-----|-------------| | **Left Arm** | 7 | Shoulder (3) + Elbow (1) + Wrist (3) | | **Right Arm** | 7 | Shoulder (3) + Elbow (1) + Wrist (3) | | **Left Hand** | 7 | Index (2) + Middle (2) + Thumb (3) | | **Right Hand** | 7 | Index (2) + Middle (2) + Thumb (3) | ## Dataset Statistics - **Trajectories:** 957 - **Total Frames:** 275,713 - **Avg Episode Length:** ~288 frames (~14.4 seconds) - **Episode Length Range:** 180-400 frames - **Storage Size:** ~1.2 GB - **Success Rate:** 100% ## Download ```bash huggingface-cli download \ --repo-type dataset jnsungp/unitree-g1-robocasa-pick-apple-bowl-1k \ --local-dir ./datasets/g1-pickplace-sim-1k ``` ### Using Python ```python from datasets import load_dataset dataset = load_dataset("jnsungp/unitree-g1-robocasa-pick-apple-bowl-1k") ``` ## Dataset Structure ``` dataset_no_modality_1k/ ├── data/ │ └── chunk-000/ │ ├── episode_000000.parquet │ ├── episode_000001.parquet │ └── ... ├── videos/ │ └── chunk-000/ │ └── observation.images.ego_view/ │ ├── episode_000000.mp4 │ ├── episode_000001.mp4 │ └── ... ├── meta/ │ ├── info.json # Dataset metadata │ ├── stats.json # Statistics (mean, std, min, max) │ ├── tasks.jsonl # Task descriptions │ └── episodes.jsonl # Episode information └── README.md ``` ## Loading Data Example ```python import pandas as pd import cv2 # Load trajectory df = pd.read_parquet("data/chunk-000/episode_000000.parquet") observations = df['observation.state'].values # (N, 28) actions = df['action'].values # (N, 28) # Load video cap = cv2.VideoCapture("videos/chunk-000/observation.images.ego_view/episode_000000.mp4") ``` ## Dataset Structure ``` ├── data/chunk-000/ # Parquet files (trajectory data) ├── videos/chunk-000/ # MP4 files (256x256, 20fps) ├── meta/ │ ├── info.json # Dataset metadata │ ├── stats.json # Statistics for normalization │ ├── tasks.jsonl # Task descriptions │ └── episodes.jsonl # Episode info ``` ## Technical Details **Simulation:** - Platform: MuJoCo + RoboCasa - Robot: Unitree G1 (upper body) - Scene: Kitchen tabletop **Motion Planning:** - CuRobo (GPU-accelerated) - Collision-free trajectories - Smooth cubic interpolation ## Citation ```bibtex @dataset{park2025unitree_g1_sim, title={Unitree G1 Apple Pick and Place Simulation Dataset}, author={Park, Junsung}, year={2025}, publisher={Hugging Face}, url={https://huggingface.co/datasets/jnsungp/unitree-g1-robocasa-pick-apple-bowl-1k} } ``` ## Acknowledgments Built with [CuRobo](https://curobo.org/), [RoboCasa](https://robocasa.ai/), [MuJoCo](https://mujoco.org/), and Unitree G1. --- **Version:** 1.0 | **Last Updated:** November 19, 2025--- 许可证：CC BY 4.0（知识共享署名4.0国际许可协议） 任务类别：机器人学 标签：unitree-g1（宇树G1）、拾取放置（pick-and-place）、仿真（simulation）、CuRobo（curobo）、运动规划（motion-planning） 数据规模类别：10万~100万样本 语言：英语 展示名称：宇树G1苹果拾取放置仿真数据集 --- # 宇树G1苹果拾取放置仿真数据集 <div align="center"> <table> <tr> <td align="center" style="vertical-align: top; padding: 5px;"> <img src="https://huggingface.co/datasets/jnsungp/unitree-g1-robocasa-pick-apple-bowl-1k/resolve/main/frontview.png" width="500px" alt="Front View"/> <br><b>全局正视图</b> </td> <td align="center" style="vertical-align: top; padding: 5px;"> <img src="https://huggingface.co/datasets/jnsungp/unitree-g1-robocasa-pick-apple-bowl-1k/resolve/main/sideview.png" width="500px" alt="Side View"/> <br><b>侧面轮廓视图</b> </td> </tr> <tr> <td align="center" style="vertical-align: top; padding: 5px;"> <img src="https://huggingface.co/datasets/jnsungp/unitree-g1-robocasa-pick-apple-bowl-1k/resolve/main/birdview.png" width="500px" alt="Top-Down View"/> <br><b>俯视鸟瞰视图</b> </td> <td align="center" style="vertical-align: top; padding: 5px;"> <img src="https://huggingface.co/datasets/jnsungp/unitree-g1-robocasa-pick-apple-bowl-1k/resolve/main/rs_view.png" width="500px" alt="Ego-Centric View"/> <br><b>机器人第一人称视角视图</b> </td> </tr> <tr> <td colspan="2" align="center" style="padding-top: 20px;"> <i>执行拾取放置任务的宇树G1机器人多视角展示。</i> </td> </tr> </table> </div> ## 数据集概述 **宇树G1苹果拾取放置仿真数据集** 包含957条高质量轨迹，记录了宇树G1人形机器人在仿真环境中完成拾取放置任务的全过程。该机器人通过双边机械臂与灵巧三指机械手，从桌面抓取红苹果并将其放置至碗中。所有轨迹均采用CuRobo运动规划生成，可实现无碰撞、平滑的运动执行。 **核心特性：** - 957条成功仿真轨迹（任务成功率100%） - 28自由度控制：双边机械臂（7+7）+ 灵巧机械手（7+7） - 20 FPS的256×256分辨率RGB第一视角视频 - 基于CuRobo的无碰撞运动规划轨迹 - 采用MuJoCo与RoboCasa仿真环境 本数据集可用于商业与非商业用途。 ## 数据集所有者 **朴俊成（Junsung Park）**（[@jnsungp](https://huggingface.co/jnsungp)） 首尔国立大学 [e-mail] ## 许可证 知识共享署名4.0国际许可协议（CC BY 4.0） ## 数据集格式 | 模态 | 数据类型 | 形状 | 描述 | |----------|------|-------|-------------| | **观测数据** | `float32` | `(28,)` | 机械臂与机械手的关节位置（弧度） | | **动作数据** | `float32` | `(28,)` | 目标关节位置 | | **视频** | RGB格式 | `(256, 256, 3)` | 第一视角视频，帧率20 FPS，编码格式H.264 | | **语言指令** | 字符串 | - | *"拿起红苹果并放置到碗中"* | ### 28自由度关节配置 | 身体部位 | 自由度数量 | 描述 | |-----------|-----|-------------| | **左臂** | 7 | 肩部（3个自由度）+ 肘部（1个自由度）+ 腕部（3个自由度） | | **右臂** | 7 | 肩部（3个自由度）+ 肘部（1个自由度）+ 腕部（3个自由度） | | **左手** | 7 | 食指（2个自由度）+ 中指（2个自由度）+ 拇指（3个自由度） | | **右手** | 7 | 食指（2个自由度）+ 中指（2个自由度）+ 拇指（3个自由度） | ## 数据集统计信息 - **轨迹总数**：957条 - **总帧数**：275,713帧 - **单段轨迹平均长度**：约288帧（约14.4秒） - **单段轨迹长度范围**：180~400帧 - **存储占用**：约1.2 GB - **任务成功率**：100% ## 下载 bash huggingface-cli download --repo-type dataset jnsungp/unitree-g1-robocasa-pick-apple-bowl-1k --local-dir ./datasets/g1-pickplace-sim-1k ### Python 使用示例 python from datasets import load_dataset dataset = load_dataset("jnsungp/unitree-g1-robocasa-pick-apple-bowl-1k") ## 数据集结构 dataset_no_modality_1k/ ├── data/ │ └── chunk-000/ │ ├── episode_000000.parquet │ ├── episode_000001.parquet │ └── ... ├── videos/ │ └── chunk-000/ │ └── observation.images.ego_view/ │ ├── episode_000000.mp4 │ ├── episode_000001.mp4 │ └── ... ├── meta/ │ ├── info.json # 数据集元数据 │ ├── stats.json # 统计信息（均值、标准差、最值） │ ├── tasks.jsonl # 任务描述文件 │ └── episodes.jsonl # 剧集信息文件 └── README.md ## 数据加载示例 python import pandas as pd import cv2 # 加载轨迹数据 df = pd.read_parquet("data/chunk-000/episode_000000.parquet") observations = df['observation.state'].values # 形状为 (N, 28) actions = df['action'].values # 形状为 (N, 28) # 加载视频文件 cap = cv2.VideoCapture("videos/chunk-000/observation.images.ego_view/episode_000000.mp4") ## 数据集结构 ├── data/chunk-000/ # Parquet 轨迹数据文件 ├── videos/chunk-000/ # MP4 视频文件（256×256分辨率，20 FPS） ├── meta/ │ ├── info.json # 数据集元数据 │ ├── stats.json # 归一化统计信息 │ ├── tasks.jsonl # 任务描述文件 │ └── episodes.jsonl # 剧集信息文件 ## 技术细节 **仿真环境：** - 平台：MuJoCo与RoboCasa - 机器人：宇树G1（上半身版本） - 场景：厨房桌面环境 **运动规划：** - 采用GPU加速的CuRobo规划器 - 生成无碰撞运动轨迹 - 使用平滑三次插值优化轨迹 ## 引用 bibtex @dataset{park2025unitree_g1_sim, title={Unitree G1 Apple Pick and Place Simulation Dataset}, author={Park, Junsung}, year={2025}, publisher={Hugging Face}, url={https://huggingface.co/datasets/jnsungp/unitree-g1-robocasa-pick-apple-bowl-1k} } ## 致谢 本数据集基于[CuRobo](https://curobo.org/)、[RoboCasa](https://robocasa.ai/)、[MuJoCo](https://mujoco.org/)以及宇树G1开发构建。 --- **版本：1.0 | 最后更新：2025年11月19日**