GR00T N1.7 AppleToPlate: 402 Real-Robot Teleoperated Demonstrations on a Unitree G1 Humanoid (NVIDIA)
- Provider
- NVIDIA Corporation
- License
- CC-BY-4.0 (source)
- Size
- 813 files
- Format
- Parquet (in sampled files)
- Stored at
- Hugging Face (https://huggingface.co/datasets/nvidia/GR00T-N1.7-AppleToPlate)
- Access
- Free and public. No account with the data provider is needed.
- Category
- Robotics
- Source
- https://huggingface.co/datasets/nvidia/GR00T-N1.7-AppleToPlate
About this dataset
Real-world humanoid robot demonstrations from NVIDIA, recorded on a physical Unitree G1 humanoid that picks up an apple and places it on a plate while standing (static loco-manipulation). All 402 trajectories were collected by human teleoperation with an XR headset at 30 Hz and were used to post-train the GR00T N1.7 vision-language-action foundation model. The data are in GR00T-LeRobot format: data/ holds one Parquet file per episode with 43-DoF whole-body joint-position actions and joint-state observations, timestamps and task indices; videos/ holds 640x480 egocentric RGB MP4 recordings; meta/ holds episodes.jsonl, tasks.jsonl, modality.json and info.json, and an embodiment configuration file for GR00T fine-tuning is included. About 1 GB (measured 2026-11-21). Read it with the lerobot library, pandas or pyarrow. Pairs with the simulated Arena G1 PickNPlace set for real-versus-sim studies of humanoid imitation learning. NVIDIA releases it under CC-BY-4.0 and states it is ready for commercial or non-commercial use; cite NVIDIA GR00T N1.
How to load this dataset in Python
from datasets import load_dataset
ds = load_dataset("nvidia/GR00T-N1.7-AppleToPlate")
print(ds)Data structure
File types in sampled files: Parquet
Files
Example file: data/chunk-000/episode_000000.parquet
Analyze this dataset without downloading it
On Scigantic, GR00T N1.7 AppleToPlate: 402 Real-Robot Teleoperated Demonstrations on a Unitree G1 Humanoid (NVIDIA) opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.
This page is also available as Markdown and as JSON.
Related Robotics datasets
- Gelsight Force Estimation
- SAGE 10 K
- L2D (Learning to Drive): Multimodal Self-Driving Dataset
- Eidon Tracker POV: Egocentric Household Video with IMU Arm Tracking
- PhysicalAI Robotics Manipulation Objects Kitchen MJCF
- IsaacLab-Arena Unitree G1 Humanoid Apple Pick-and-Place: 200 XR-Teleoperated Simulation Demonstrations (NVIDIA)
- GrandTour: Legged Robot Multi-Sensor Dataset (ETH Zurich)
- PhysicalAI Traffic Anomaly Reasoning