Geppetto tracks 1035 robot training datasets including Open X-Embodiment, DROID, and AgiBot World, each with structured metadata on trajectory count, hours of demonstration, robot platforms, license, commercial use, access model, and file format. Every dataset connects to the robots it trains and the professions it affects.
| Dataset | Year | Category | Trajectories | Hours | Platforms | License | Commercial | Access | Format |
|---|---|---|---|---|---|---|---|---|---|
| NavWareSet | 2026 | research-lab | — | — | mobile robot | not stated | No | open | — |
| STONE Dataset | 2026 | research-lab | — | — | off-road robot, surround-view multi-modal | not stated | No | open | — |
| Wall-Following Robot Navigation Data | 2026 | research-lab | — | — | mobile robot ultrasound sensors | not stated | No | open | — |
| DARP | 2026 | research-lab | — | — | dual-arm RGB-D-IR rig | not stated | No | open | — |
| 2026-challenge-demos | 2026 | synthetic | 20,000 | 1,950 | JoyLo | MIT | Yes | open | — |
| FineART | 2026 | real-robot | 40,543 | 1,718 | Trossen Stationary AI | CC-BY-4.0 | Yes | open | — |
| PhysicalAI Robotics GR00T Teleop G1 | 2025 | real-robot | — | — | Unitree G1 | CC BY 4.0 | Yes | open | parquet, hdf5, mp4 |
| Human2LocoMan — Loco-Manipulation Dataset | 2025 | real-robot | — | — | LocoMan | MIT | Yes | open | hdf5, mp4 |
| RealSource-World — Dual-Arm Humanoid Manipulation Dataset | 2025 | real-robot | — | — | RS-02 dual-arm humanoid | CC BY-NC-SA 4.0 — non-commercial use only | No | gated | parquet, mp4 |
| DexWild — In-the-Wild Dexterous Manipulation Dataset | 2025 | human-video | — | — | 6 DoF dexterous hand | MIT | Yes | open | imagefolder, parquet |
| EgoVerse — Egocentric Robot Training Dataset: 8.5M Frames, 74 Countries | 2025 | real-robot | — | 1,213 | aria_bimanual, aria_left_arm, aria_right_arm | CC BY-SA 4.0 | Yes | open | zarr, lerobot, hdf5 |
| AgiBot-Alpha — AgiBot World Dataset | 2025 | real-robot | — | 1,979 | AgiBot A2, AgiBot X1 | AgiBot World Open Academic License — commercial use with restrictions | Yes | open | zarr, mp4 |
| LeRobot Community Datasets v3 | 2025 | real-robot | — | — | SO-100 arm, SO-101 arm, Franka Panda | Apache 2.0 | Yes | open | parquet, mp4, lerobot |
| OXE-AugE | 2025 | synthetic | 4,440,000 | — | Franka Emika Panda, UR5e, xArm7 | CC-BY-4.0 | Yes | open | — |
| RFUAV | 2025 | research-lab | — | — | USRP radio frequency recordings of 37 UAVs | not stated | No | open | — |
| Bench-RNR | 2025 | research-lab | — | — | infrastructure LiDARs, repetitive and non-repetitive scanning | not stated | No | open | — |
| CoInfra | 2025 | research-lab | — | — | 8-node roadside lidar and camera system at roundabout | not stated | No | open | — |
| R-LiViT | 2025 | research-lab | — | — | roadside lidar, RGB and thermal cameras | not stated | No | open | — |
| REHEARSE-3D | 2025 | research-lab | — | — | LiDAR and radar in emulated rain | not stated | No | open | — |
| DISC | 2025 | research-lab | — | — | simulated vehicle | not stated | No | open | — |
| SimCoachCorpus | 2025 | research-lab | — | — | driving simulator, race track | not stated | No | open | — |
| AevaScenes | 2025 | research-lab | — | — | Aeva FMCW lidar vehicle | not stated | No | open | — |
| AGC-Drive | 2025 | research-lab | — | — | vehicles plus UAV | not stated | No | open | — |
| CATS-V2V | 2025 | research-lab | — | — | two vehicles with lidar and multi-view cameras | not stated | No | open | — |
| Dur360BEV | 2025 | research-lab | — | — | car with spherical camera, 128-channel lidar, RTK GNSS/INS | not stated | No | open | — |
| EMT | 2025 | research-lab | — | — | dash camera, UAE | not stated | No | open | — |
| MS2 (Multi-Spectral Stereo) | 2025 | research-lab | — | — | stereo RGB, NIR, thermal, LiDAR | not stated | No | open | — |
| NavINST | 2025 | research-lab | — | — | multi-sensor road test vehicle | not stated | No | open | — |
| Odyssey | 2025 | research-lab | — | — | automotive LiDAR-inertial rig | not stated | No | open | — |
| OnSiteVRU | 2025 | research-lab | — | — | vulnerable road user trajectories | not stated | No | open | — |
| RASMD | 2025 | research-lab | — | — | RGB and SWIR cameras | not stated | No | open | — |
| RISEE | 2025 | research-lab | — | — | naturalistic trajectories with risk ratings | not stated | No | open | — |
| RoadSens-4M | 2025 | research-lab | — | — | smartphone sensors and camera | not stated | No | open | — |
| RoRaTrack | 2025 | research-lab | — | — | Dallara AV-21 camera | not stated | No | open | — |
| STU (Spotting the Unexpected) | 2025 | research-lab | — | — | lidar and camera | not stated | No | open | — |
| TAD-E2E | 2025 | research-lab | — | — | not stated on page opened | not stated | No | open | — |
| UrbanIng-V2X | 2025 | research-lab | — | — | multiple vehicles and infrastructure across intersections | not stated | No | open | — |
| UrbanV2X | 2025 | research-lab | — | — | vehicle and infrastructure sensors, C-V2X | not stated | No | open | — |
| V2X-ReaLO | 2025 | research-lab | — | — | V2X-Real vehicles and infrastructure, ROS bags | not stated | No | open | — |
| Waymo internal driving dataset (Scaling Laws of Motion Forecasting and Planning) | 2025 | research-lab | — | — | not stated | not stated | No | open | — |
| DexUMI | 2025 | human-video | — | — | wearable hand exoskeleton | not stated | No | open | — |
| Aria Gen 2 Pilot Dataset | 2025 | human-video | — | — | Project Aria Gen 2 glasses | Project Aria dataset terms (per atlas) | No | open | — |
| Aria STDR | 2025 | human-video | — | — | Project Aria glasses | not stated | No | open | — |
| Digital Twin Catalog (DTC) | 2025 | human-video | — | — | Project Aria glasses and DSLR | dataset licence agreement | No | open | — |
| ECBench | 2025 | human-video | — | — | egocentric videos | not stated | No | open | — |
| Ego-HOIBench | 2025 | human-video | — | — | Egocentric images with hand-verb-object triplets | not stated | No | open | — |
| EgoBlind | 2025 | human-video | — | — | cameras used by blind people | not stated | No | open | — |
| EgoCross | 2025 | human-video | — | — | egocentric video from surgery, industry, extreme sports, animal views | not stated | No | open | — |
| EgoEdit (EgoEditData) | 2025 | human-video | — | — | egocentric videos | not stated | No | gated | — |
| EgoEvGesture | 2025 | human-video | — | — | egocentric event camera | not stated | No | open | — |
| EgoExo-Con | 2025 | human-video | — | — | synchronised ego-exo video pairs | not stated | No | open | — |
| EgoGazeVQA | 2025 | human-video | — | — | Egocentric daily-life video with gaze | not stated | No | open | — |
| EgoIllusion | 2025 | human-video | — | — | egocentric videos | not stated | No | open | — |
| EgoNight | 2025 | human-video | — | — | egocentric video, day and night pairs | not stated | No | open | — |
| EgoNormia | 2025 | human-video | — | — | egocentric videos of human interactions | not stated | No | open | — |
| EgoPrivacy | 2025 | human-video | — | — | existing egocentric videos | not stated | No | open | — |
| EgoRe-5M (EgoThinker) | 2025 | human-video | — | — | Egocentric video clips from existing sources | not stated | No | open | — |
| EgoSocial | 2025 | human-video | — | — | Egocentric social interaction video | not stated | No | open | — |
| EgoTempo | 2025 | human-video | — | — | egocentric videos | not stated | No | open | — |
| EgoTextVQA | 2025 | human-video | — | — | ego-view videos, driving and house-keeping | not stated | No | open | — |
| EgoToM | 2025 | human-video | — | — | Ego4D videos | not stated | No | open | — |
| EgoVLA training mix and Ego Humanoid Manipulation Benchmark | 2025 | human-video | — | — | egocentric human videos; sim benchmark uses Unitree H1 with Inspire hands | MIT (Hub EgoVLA/EgoVLA-Humanoid-Sim) | Yes | open | — |
| EMMA | 2025 | human-video | — | — | Project Aria glasses | not stated | No | open | — |
| EOC-Bench | 2025 | human-video | — | — | Egocentric video | not stated | No | open | — |
| EvHand-FPV | 2025 | human-video | — | — | Single event camera, first-person view | not stated | No | open | — |
| ExAct | 2025 | human-video | — | — | skilled-activity ego and exo videos | not stated | No | open | — |
| GuideDog | 2025 | human-video | — | — | pedestrian first-person images | not stated | No | open | — |
| LSC-ADL | 2025 | human-video | — | — | lifelog wearable camera images (LSC) | not stated | No | open | — |
| TASTE-Rob | 2025 | human-video | — | — | fixed-view ego-centric camera | not stated | No | open | — |
| TeleEgo | 2025 | human-video | — | — | wearable camera with audio | not stated | No | open | — |
| UniHand (Being-H0) | 2025 | human-video | — | — | mixed sources: motion capture, VR headsets and RGB-only videos | not stated | No | open | — |
| DG16M | 2025 | research-lab | — | — | dual-arm grasps | not stated | No | open | — |
| Dexonomy | 2025 | research-lab | — | — | dexterous hands, 31 GRASP taxonomy types | not stated | No | open | — |
| Learning Gentle Grasping dataset | 2025 | research-lab | — | — | multi-fingered hand with vision, sound and touch | not stated | No | open | — |
| ScaleADFG | 2025 | research-lab | — | — | 3D assets with dexterous functional grasps | not stated | No | open | — |
| UniFucGrasp | 2025 | research-lab | — | — | multiple dexterous hand types | not stated | No | open | — |
| SpikeGrasp | 2025 | research-lab | — | — | stereo spike cameras | not stated | No | open | — |
| TD-TOG | 2025 | research-lab | — | — | real RGB-D scenes with affordances and planar grasps | not stated | No | open | — |
| GraspFactory | 2025 | research-lab | — | — | Franka Panda and Robotiq 2F-85 grippers | custom licence file | No | open | — |
| LIBERO-Plus | 2025 | synthetic | — | 31.08 | franka_panda | MIT | Yes | open | — |
| Humanoid Everyday | 2025 | real-robot | — | 27.8 | unitree_humanoid | Apache-2.0 | Yes | open | — |
| AIRoA MoMa Dataset | 2025 | real-robot | — | 94 | — | custom | No | open | — |
| U2UData+ | 2025 | synthetic | — | 120 | drone | Apache-2.0 | Yes | open | — |
| RoboArena | 2025 | real-robot | — | — | franka_panda, parallel_jaw | MIT | Yes | open | — |
| EgoDex | 2025 | human-video | — | 829 | human | research-only | No | open | — |
| UAV-Flow-Sim | 2025 | synthetic | — | — | drone | unknown | No | open | — |
| UAV-Flow | 2025 | human-video | — | — | drone | unknown | No | open | — |
| CognitiveDrone Dataset | 2025 | synthetic | — | — | drone | unknown | No | open | — |
| VideoPhy-2 | 2025 | research-lab | — | — | generated action-centric video | MIT | Yes | open | — |
| 1000 Robot Manipulation Tasks | 2025 | research-lab | — | — | not stated | mit | Yes | open | — |
| X-Capture | 2025 | research-lab | — | — | portable device: RGB-D, tactile, audio, 3D | MIT | Yes | open | — |
| OpenLoong mocap data | 2025 | research-lab | — | — | human BVH skeleton for OpenLoong humanoid | Apache 2.0 | Yes | open | — |
| Unitree Z1 Dual-arm datasets | 2025 | real-robot | — | — | Unitree Z1 dual arm with Dex1 grippers | Apache-2.0 | Yes | open | — |
| FeelAnyForce | 2025 | research-lab | — | — | not stated on card | CC BY 4.0 | Yes | open | — |
| VitaSet | 2025 | research-lab | — | — | RGB vision and tactile sensing via robotic manipulation | MIT | Yes | open | — |
| Wanderland | 2025 | research-lab | — | — | not stated | apache-2.0 | Yes | open | — |
| PhysicalAI-Robotics-GR00T-GR1 (GR1-100) | 2025 | real-robot | — | — | Fourier GR1-T2 (real), third-person video only | CC BY 4.0 | Yes | open | — |
| WorldScore | 2025 | research-lab | — | — | static and dynamic scenes with camera trajectories | MIT | Yes | open | — |
| ManiFeel | 2025 | synthetic | — | — | visuotactile manipulation benchmark | MIT | Yes | open | — |
| Mixed Signals | 2025 | research-lab | — | — | 3 connected vehicles plus roadside unit, 5 lidars | CC BY-NC-SA 4.0 | No | open | — |