← Sinan Lab › Public robot datasets
Public robot datasets
The other half of the model index: what data exists to fine-tune a model to your robot. Scale, embodiments, format, licence and China mirrors, each field with a source. Only what the dataset page or paper states; blanks are unknowns, not zeros.
32 datasets13,028 hours stated in total7,572,222 episodes stated in total
32 datasets match
| Dataset | Embodiments | Scale | Format | Licence | Mirror |
|---|---|---|---|---|---|
| InternData-A1 InternRobotics · 2025-11 | dual armsingle armmulti-embodimentsimulation ARX Lift-2 · AgileX Split Aloha · A2D · Franka | 7,433 h 630,000 episodes | LeRobot v2.1 and LeRobot v… | ✗ CC BY-NC-SA 4.0 (HF card); ModelScope metadata l | ModelScope |
| Ego4D Ego4D consortium (licenses issued per partner institution, e.g. University of Bristol) · 2022-02 | human video human (head-mounted cameras: GoPro, Vuzix Blade, Pupil Labs, ZShades, OR-DRO EP6, iVue Rincon 1080, Weeview) | 3,670 h | custom (canonical MP4 vide… | ✓ Ego4D License Agreement (signed per-institution | — |
| EgoDex Apple · 2025-05 | human handhuman video human hands (egocentric video, Apple Vision Pro) | 829 h 338,000 episodes | HDF5 + MP4 (paired per epi… | ✗ CC-by-NC-ND terms | — |
| RoboCasa (RoboCasa365) robocasa.ai | simulationmobile manipulatorhumanoidquadrupedmulti-embodiment mobile manipulators of diverse form factors, such as single-arm mobile platforms, humanoid robots, and quadruped robots with arms | 600 h | ? | ✓ Code: MIT License; Assets/Datasets: CC BY 4.0 | — |
| DROID (Distributed Robot Interaction Dataset) DROID collaboration (13 institutions) · 2024-03 | single arm Franka Panda 7DoF robot arm | 350 h 76,000 episodes | RLDS/TFDS | ? | — |
| Fourier ActionNet Fourier Intelligence (FFTAI) | humanoiddual arm Fourier GR1-T1 · Fourier GR1-T2 · Fourier GR2 · Fourier Dexterous hand (6 DoF) … | 140 h 30,000 episodes | LeRobot v2 (LeRobotDataset… | ✗ CC BY-NC-SA 4.0 | — |
| Kaiwu Tongji University (Shuo Jiang, Haonan Li, Ruochen Ren, Yanmin Zhou, Zhipeng Wang, Bin He) · 2025-03 | human handhuman video human (20 subjects; data glove, EMG, motion capture, eye tracking) | 6 h 11,664 episodes | custom (raw data directori… | ? | — |
| ARIO (All Robots In One) ARIO team (hosted on OpenI, Peng Cheng Laboratory) · 2024-08 | dual armmobile manipulatorwheeledhumanoidsingle armmulti-embodimentsimulation Cobot Magic (AgileX Robotics, bimanual mobile manipulation platform) · Cloud Ginger XR-1 (wheeled humanoid cloud robot) · UR5 (MuJoCo simulation) · embodiments of converted open-source datasets (Open X-Embodiment, RH20T) | 3,033,188 episodes | custom (ARIO unified forma… | ? CC-BY 4.0 or MIT for ARIO-collected real data an | — |
| AgiBot World (Alpha / Beta) AgiBot (agibot-world) / OpenDriveLab · 2024-12 | dual armmobile manipulator mobile dual-arm robots · 6-DoF Dexterous hand | 1,003,672 episodes | LeRobot (README); HF card … | ✗ CC BY-NC-SA 4.0 | — |
| Open X-Embodiment (OXE) Open X-Embodiment Collaboration (Google DeepMind + partner labs) · 2023-10 | multi-embodiment 22 different robots | 1,000,000 episodes | RLDS | ? Software: Apache License 2.0; all other material | — |
| RoboVerse (roboverse_data) RoboVerseOrg · 2025-04 | single armhumanoidmulti-embodimentsimulation Franka | 500,000 episodes | ? | ? apache-2.0 (HF license tag); paper states 'CC BY | — |
| PhysicalAI-Robotics-GR00T-X-Embodiment-Sim NVIDIA | dual armhumanoidsingle armmulti-embodimentsimulation bimanual Panda (gripper) · bimanual Panda (hand) · GR1 (arms only) · GR1 (full upper body) … | 345,102 episodes | LeRobot v2 layout (GR00T L… | ✓ cc-by-4.0 | — |
| RT-1 dataset (Google robot) Google · 2022-12 | mobile manipulatorsingle arm fleet of 13 robots (platform not named on project page) | 130,000 episodes | ? | ? | — |
| RH20T RH20T authors (rh20t.github.io) · 2023-07 | multi-embodimenthuman video 7 robot configurations (cfg1-cfg7; models detailed in paper appendix) | 110,000 episodes | custom (MP4 video + NPY ar… | ? RH20T-C subset: CC BY-SA 4.0; RH20T-NC subset: C | — |
| RoboMIND X-Humanoid (x-humanoid-robomind) · 2024-12 | single armdual armhumanoidsimulationmulti-embodiment Franka Emika Panda · UR5e · AgileX dual-arm robot · humanoid robot with dual dexterous hands … | 107,000 episodes | ? | ✓ apache-2.0 | — |
| RoboTwin 2.0 RoboTwin-Platform (Tianxing Chen et al.) · 2025-06 | dual armsimulation aloha_agilex (default embodiment) | 100,000 episodes | HDF5 | ✓ MIT | — |
| BridgeData V2 UC Berkeley RAIL · 2023-08 | single arm WidowX 250 6DOF robot arm | 60,096 episodes | raw (JPEG images, zip down… | ✓ Creative Commons Attribution 4.0 International L | — |
| MimicGen Datasets NVIDIA (NVlabs) · 2023-10 | single armsimulation Panda · Sawyer · IIWA · UR5e | 48,000 episodes | HDF5 (robomimic) | ✓ CC-BY 4.0 | — |
| RoboSet RoboAgent / RoboPen (robopen.github.io) · 2023-09 | 28,500 episodes | HDF5 (h5py) | ✓ MIT-License | — | |
| DexMimicGen Datasets NVIDIA (NVlabs) · 2024-10 | dual armhumanoidsimulation bimanual Panda arms with parallel-jaw grippers · bimanual Panda arms with dexterous hands · GR-1 humanoid with dexterous hands | 21,000 episodes | HDF5 | ? cc-by-nc-sa-4.0 (HF card YAML); GitHub README st | — |
| LIBERO datasets Lifelong-Robot-Learning (LIBERO project) · 2023-06 | simulation | ? | ? | ✓ Codebase: MIT License; Datasets: Creative Common | — |
| CALVIN mees/calvin (Oier Mees et al.) · 2021-12 | simulationsingle arm | ? | ? | ✓ MIT | — |
| Galaxea Open-World Dataset (G0 data) Galaxea (OpenGalaxea) · 2025-08 | dual armmobile manipulator Galaxea R1 Lite (mobile dual-arm manipulator) | ? | LeRobot (HF tag 'lerobot') | ✗ cc-by-nc-sa-4.0 | — |
| BC-Z dataset Google · 2022-02 | ? | ? | ? | — | |
| UniHand (Being-H0 human hand pretraining data; HF: UniHand_Preview) BeingBeyond · 2025-07 | human handhuman video human hands (motion capture, VR, and RGB-only videos) | ? | webdataset (HF card format… | ? | — |
| LeRobot community datasets (SO-100 / SO-101) Hugging Face LeRobot community | so100 · so101 | ? | ? | ? | — |
| ManiSkill3 Demonstrations haosulab (ManiSkill team) · 2024-10 | single armmobile manipulatorhumanoidquadrupedmulti-embodimentsimulation Anymal C · Allegro Hand Left · Allegro Hand Right · Allegro Hand Right with Touch Sensing … | ? | HDF5 + JSON metadata | ? Apache-2.0 (GitHub repository LICENSE); dataset | — |
| RLBench Imperial College London (Stephen James et al.) · 2019-09 | single armsimulation Franka Panda Arm (default) · Mico arm with Mico gripper · Jaco arm with 3-finger Jaco gripper · Sawyer arm with Baxter gripper … | ? | custom (CoppeliaSim/V-REP … | ✗ Custom Imperial College London license (non-comm | — |
| Unitree Robotics open datasets (Hugging Face org) Unitree Robotics · 2025-01 | humanoiddual armsingle armmulti-embodiment Unitree G1 (robot_type 'unitree_g1') · G1 with Dex1 gripper (dataset ids G1_Dex1_*) · G1 with Dex3 hand (G1_Dex3_*) · G1 with Inspire hand (G1_WBT_Inspire_*) … | ? | LeRobot v3 (codebase_versi… | ✓ Apache-2.0 (94 of 160 datasets tagged license:ap | — |
| AgiBot DigitalWorld (+ AgiBot World Challenge datasets) AgiBot (AgiBot World / OpenDriveLab) · 2025-02 | dual armsimulation AgiBot dual-arm robot, 34 DoF (2 DoF head, 2 DoF waist, 7 DoF per arm, 8 DoF per gripper) | ? | custom (MP4 RGB + PNG dept… | ✗ CC BY-NC-SA 4.0 | ModelScope |
| UMI (Universal Manipulation Interface) datasets Stanford REAL Lab (real-stanford) / Columbia · 2024-02 | single arm hand-held parallel jaw gripper with GoPro (data collection device) · UR5e · UR5 · Franka … | ? | zarr (.zarr.zip) | ? MIT (code repository); dataset license not separ | — |
| Open-TeleVision teleoperation datasets OpenTeleVision (Xuxin Cheng et al.) · 2024-07 | humanoiddual arm Unitree H1 with 6 DoF Inspire hands · Fourier GR-1 with jaw gripper | ? | custom (raw recordings pla… | ? Apache-2.0 (code repository); dataset license no | — |
Sorted by stated hours, then episodes. Simulation datasets are tagged “simulation”. Licence classification is literal: CC-BY / Apache / MIT → commercial OK, CC-BY-NC → not, anything else → unknown.
Clear direction, grounded judgement.