diff --git a/docs/registry.html b/docs/registry.html index 577ddee..4684f6b 100644 --- a/docs/registry.html +++ b/docs/registry.html @@ -316,6 +316,7 @@ .badge-demo { color: var(--orange); background: var(--orange-dim); border-color: rgba(251,191,36,0.2); } .badge-tag { color: var(--text-secondary); background: rgba(94,110,138,0.1); border-color: rgba(94,110,138,0.15); } .badge-scale { color: var(--teal); background: var(--teal-dim); border-color: rgba(45,212,191,0.2); } + .badge-gated { color: var(--rose); background: var(--rose-dim); border-color: rgba(251,113,133,0.25); } .chevron { color: var(--text-muted); @@ -521,7 +522,7 @@

Dataset Registry

⚒ Dataset facts labels →
-
23
+
32
Datasets
@@ -529,15 +530,15 @@

Dataset Registry

Demo-Ready
-
3.6M
+
5.0M
Total Episodes
-
5
+
6
Formats
-
12
+
16
Robots
@@ -557,6 +558,7 @@

Dataset Registry

+ + + + + + - 23 datasets + 32 datasets
-
+
@@ -614,6 +626,8 @@

Dataset Registry

large + +
@@ -727,7 +741,7 @@

Dataset Registry

-
+
@@ -742,6 +756,8 @@

Dataset Registry

large + +
@@ -860,7 +876,7 @@

Dataset Registry

-
+
@@ -875,6 +891,8 @@

Dataset Registry

large + +
@@ -993,7 +1011,7 @@

Dataset Registry

-
+
@@ -1008,6 +1026,8 @@

Dataset Registry

large + +
@@ -1121,7 +1141,7 @@

Dataset Registry

-
+
@@ -1136,6 +1156,8 @@

Dataset Registry

+ +
@@ -1248,7 +1270,7 @@

Dataset Registry

-
+
@@ -1263,6 +1285,8 @@

Dataset Registry

+ +
@@ -1375,7 +1399,7 @@

Dataset Registry

-
+
@@ -1390,6 +1414,8 @@

Dataset Registry

+ +
@@ -1502,7 +1528,7 @@

Dataset Registry

-
+
@@ -1517,6 +1543,8 @@

Dataset Registry

+ +
@@ -1629,7 +1657,7 @@

Dataset Registry

-
+
@@ -1644,6 +1672,8 @@

Dataset Registry

large + +
@@ -1753,7 +1783,7 @@

Dataset Registry

-
+
@@ -1768,6 +1798,8 @@

Dataset Registry

large + +
@@ -1872,7 +1904,7 @@

Dataset Registry

-
+
@@ -1887,6 +1919,8 @@

Dataset Registry

large + +
@@ -1991,7 +2025,7 @@

Dataset Registry

-
+
@@ -2006,6 +2040,8 @@

Dataset Registry

+ +
@@ -2115,7 +2151,7 @@

Dataset Registry

-
+
@@ -2130,6 +2166,8 @@

Dataset Registry

+ +
@@ -2239,7 +2277,7 @@

Dataset Registry

-
+
@@ -2254,6 +2292,8 @@

Dataset Registry

+ +
@@ -2358,7 +2398,7 @@

Dataset Registry

-
+
@@ -2373,6 +2413,8 @@

Dataset Registry

large + +
@@ -2477,7 +2519,7 @@

Dataset Registry

-
+
@@ -2492,6 +2534,8 @@

Dataset Registry

large + +
@@ -2601,7 +2645,7 @@

Dataset Registry

-
+
@@ -2616,6 +2660,8 @@

Dataset Registry

+ +
@@ -2720,7 +2766,7 @@

Dataset Registry

-
+
@@ -2735,6 +2781,8 @@

Dataset Registry

+ +
@@ -2853,7 +2901,7 @@

Dataset Registry

-
+
@@ -2868,6 +2916,8 @@

Dataset Registry

+ +
@@ -2986,7 +3036,7 @@

Dataset Registry

-
+
@@ -2999,6 +3049,8 @@

Dataset Registry

franka + +
@@ -3090,7 +3142,7 @@

Dataset Registry

-
+
@@ -3105,6 +3157,8 @@

Dataset Registry

+ +
@@ -3214,7 +3268,7 @@

Dataset Registry

-
+
@@ -3229,6 +3283,8 @@

Dataset Registry

+ +
@@ -3341,7 +3397,7 @@

Dataset Registry

-
+
@@ -3356,6 +3412,8 @@

Dataset Registry

+ +
@@ -3468,6 +3526,1169 @@

Dataset Registry

+
+
+
+
+ RoboCasa + robocasa +
+
Large-scale simulated kitchen manipulation: ~100 human demos per task across 100+ atomic/composite tasks, expanded to ~100k+ trajectories with MimicGen in MuJoCo. A standard pretraining corpus behind GR00T.
+
+ hdf5 + franka + + + large + + + +
+
+ +
+ +
+ + + + 100,000 episodes + + + + + + 600h + + + + + + MIT + + +
+ +
+
+
+ + +
+
Paper
+ + + arXiv:2406.02523 + +
+ + + +
+
Tags
+
+ manipulationsimulationcontact richmulti tasklarge scale +
+
+ + + +
+
Task Types
+
+ pick placedrawerpushcontact richmulti task +
+
+ + +
+
Sources
+
+ +
+ http + + https://robocasa.ai/ + + Canonical download via robocasa download_datasets.py (robomimic-style HDF5) + +
+ +
+ hf_hub + + haosulab/RoboCasa + + ManiSkill/SAPIEN-adapted mirror; LeRobot-converted versions also on the Hub + +
+ +
+
+ +
+
Forge Commands
+
+
+ forge inspect robocasa + +
+
+ forge quality robocasa + +
+
+ forge convert robocasa ./output --format lerobot-v3 + +
+ +
+
+ + +
+
Notes
+
MimicGen-generated + human demos; kitchen-centric. LeRobot conversions widely available for easy ingest.
+
+ + +
+
+
+
+ +
+
+
+
+ ManiSkill (v3) + maniskill +
+
GPU-parallelized manipulation simulator and benchmark with millions of demonstration frames across diverse task families (pick-place, stacking, articulated objects, dexterous).
+
+ hdf5 + frankawidowxdexterous_hand + + + + + + +
+
+ +
+ +
+ + + + 30,000 episodes + + + + + + + Apache-2.0 + + +
+ +
+
+
+ + +
+
Paper
+ + + arXiv:2410.00425 + +
+ + + +
+
Tags
+
+ manipulationsimulationcontact richmulti task +
+
+ + + +
+
Task Types
+
+ pick placepushdrawercontact richmulti task +
+
+ + +
+
Sources
+
+ +
+ hf_hub + + haosulab/ManiSkill_Demonstrations + + Download per-task via `python -m mani_skill.utils.download_demo <env_id>` + +
+ +
+
+ +
+
Forge Commands
+
+
+ forge inspect maniskill + +
+
+ forge quality maniskill + +
+
+ forge convert maniskill ./output --format lerobot-v3 + +
+ +
+
+ + +
+
Notes
+
Widely-used sim benchmark; trajectory .h5 + json. Episode count varies by task family.
+
+ + +
+
+
+
+ +
+
+
+
+ CALVIN + calvin +
+
Long-horizon, language-conditioned tabletop manipulation benchmark across four env splits (A/B/C/D); ~2M teleoperated trajectories with ~24k language-annotated.
+
+ other + franka + + + + + + +
+
+ +
+ +
+ + + + 24,000 episodes + + + + + + 24h + + + + + + MIT + + +
+ +
+
+
+ + +
+
Paper
+ + + arXiv:2112.03227 + +
+ + + +
+
Tags
+
+ manipulationsimulationlanguage conditionedmulti task +
+
+ + + +
+
Task Types
+
+ pick placedrawerpushmulti task +
+
+ + +
+
Sources
+
+ +
+ http + + https://github.com/mees/calvin + + Download via dataset/download_data.sh (D, ABC, or ABCD splits); custom .npz frames + +
+ +
+
+ +
+
Forge Commands
+
+
+ forge inspect calvin + +
+
+ forge quality calvin + +
+
+ forge convert calvin ./output --format lerobot-v3 + +
+ +
+
+ + +
+
Notes
+
Reference long-horizon language-conditioned benchmark; ~24k of ~2M trajectories carry language labels.
+
+ + +
+
+
+
+ +
+
+
+
+ robomimic + robomimic +
+
Canonical imitation-learning / offline-RL manipulation benchmark: proficient- and mixed-human teleop demonstrations across sim and real tasks (Lift, Can, Square, Transport, Tool Hang).
+
+ hdf5 + franka + + + + + + +
+
+ +
+ +
+ + + + 6,000 episodes + + + + + + + MIT + + +
+ +
+
+
+ + +
+
Paper
+ + + arXiv:2108.03298 + +
+ + + +
+
Tags
+
+ manipulationsimulationcontact richmulti task +
+
+ + + +
+
Task Types
+
+ pick placecontact richmulti task +
+
+ + +
+
Sources
+
+ +
+ http + + https://robomimic.github.io/docs/datasets/robomimic_v0.1.html + + Download via download_datasets.py; native robomimic HDF5 + +
+ +
+
+ +
+
Forge Commands
+
+
+ forge inspect robomimic + +
+
+ forge quality robomimic + +
+
+ forge convert robomimic ./output --format lerobot-v3 + +
+ +
+
+ + +
+
Notes
+
The reference HDF5 layout much of the ecosystem (incl. MimicGen, RoboCasa) inherits.
+
+ + +
+
+
+
+ +
+
+
+
+ MimicGen + mimicgen +
+
Auto-generated manipulation demonstrations: ~48k trajectories across 12 tasks synthesized from ~200 human demos, with diverse object/scene configs and multiple arms.
+
+ hdf5 + frankasawyerur5 + + + + + + +
+
+ +
+ +
+ + + + 48,000 episodes + + + + + + + CC-BY-4.0 + + +
+ +
+
+
+ + +
+
Paper
+ + + arXiv:2310.17596 + +
+ + + +
+
Tags
+
+ manipulationsimulationcontact richmulti tasklarge scale +
+
+ + + +
+
Task Types
+
+ pick placecontact richmulti task +
+
+ + +
+
Sources
+
+ +
+ hf_hub + + amandlek/mimicgen_datasets + + robomimic-format HDF5 (source/core/object groups) + +
+ +
+
+ +
+
Forge Commands
+
+
+ forge inspect mimicgen + +
+
+ forge quality mimicgen + +
+
+ forge convert mimicgen ./output --format lerobot-v3 + +
+ +
+
+ + +
+
Notes
+
Data-generation companion to robomimic; datasets CC-BY-4.0 (code is NVIDIA Source Code License).
+
+ + +
+
+
+
+ +
+
+
+
+ RoboMIND + robomind +
+
Multi-embodiment real-world teleoperation dataset: ~107k trajectories, 479 tasks, 96 object classes across Franka single-arm, AgileX dual-arm, UR-5e, and the Tien Kung humanoid.
+
+ hdf5 + frankaur5agilextien_kung + + + large + + 🔒 registration + +
+
+ +
+ +
+ + + + 107,000 episodes + + + + + + 1400h + + + + + + Apache-2.0 + + +
+ +
+
+
+ + +
+
Paper
+ + + arXiv:2412.13877 + +
+ + + +
+
Tags
+
+ manipulationbi manualhumanoidreal worldmulti tasklarge scaleregistration required +
+
+ + + +
+
Task Types
+
+ pick placepushdrawermulti task +
+
+ + +
+
Sources
+
+ +
+ hf_hub + + x-humanoid-robomind/RoboMIND + + Requires platform registration/agreement; LeRobot-format versions available + +
+ +
+
+ +
+
Forge Commands
+
+
+ forge inspect robomind + +
+
+ forge quality robomind + +
+
+ forge convert robomind ./output --format lerobot-v3 + +
+ +
+
+ + +
+
Notes
+
Access requires registration (BAAI/HF gated). Strong multi-embodiment + humanoid coverage.
+
+ + +
+
+
+
+ +
+
+
+
+ AgiBot World Beta + agibot_world_beta +
+
Large-scale real-world dual-arm humanoid manipulation: 1M+ trajectories, ~2,976 hours, 217 tasks and 87 skills with dexterous hands and mobile bases. Supersedes AgiBot World Alpha.
+
+ lerobot-v2 + agibot_g1 + + + large + + + ⊘ non-commercial +
+
+ +
+ +
+ + + + 1,000,000 episodes + + + + + + 2976h + + + + + + CC-BY-NC-SA-4.0 + + +
+ +
+
+
+ + +
+
Paper
+ + + arXiv:2503.06669 + +
+ + + +
+
Tags
+
+ manipulationbi manualhumanoidreal worldmulti tasklarge scalenon commercial +
+
+ + + +
+
Task Types
+
+ pick placepourcontact richmulti task +
+
+ + +
+
Sources
+
+ +
+ hf_hub + + agibot-world/AgiBotWorld-Beta + + + +
+ +
+
+ +
+
Forge Commands
+
+
+ forge inspect agibot_world_beta + +
+
+ forge quality agibot_world_beta + +
+
+ forge convert agibot_world_beta ./output --format lerobot-v3 + +
+ +
+
+ + +
+
Notes
+
Non-commercial license (CC-BY-NC-SA-4.0). Full-scale successor to agibot_world_alpha.
+
+ + +
+
+
+
+ +
+
+
+
+ RoboTwin 2.0 + robotwin +
+
Scalable bimanual manipulation data generator and benchmark: 100k+ domain-randomized expert trajectories across 50 dual-arm tasks and five embodiments (incl. Aloha-AgileX).
+
+ hdf5 + alohafrankaur5 + + + large + + + +
+
+ +
+ +
+ + + + 100,000 episodes + + + + + + + Apache-2.0 + + +
+ +
+
+
+ + +
+
Paper
+ + + arXiv:2506.18088 + +
+ + + +
+
Tags
+
+ manipulationbi manualsimulationcontact richmulti tasklarge scale +
+
+ + + +
+
Task Types
+
+ pick placecontact richmulti task +
+
+ + +
+
Sources
+
+ +
+ hf_hub + + TianxingChen/RoboTwin2.0 + + RoboTwin2_dataset (~80GB); LeRobot-compatible + +
+ +
+
+ +
+
Forge Commands
+
+
+ forge inspect robotwin + +
+
+ forge quality robotwin + +
+
+ forge convert robotwin ./output --format lerobot-v3 + +
+ +
+
+ + +
+
Notes
+
Strong domain randomization; five arms (Franka, Piper, UR5, ARX-X5, Aloha-AgileX).
+
+ + +
+
+
+
+ +
+
+
+
+ UMI (Universal Manipulation Interface) + umi +
+
In-the-wild human demonstrations collected with handheld grippers, yielding hardware-agnostic, deployable manipulation policies (incl. bimanual and dynamic tasks).
+
+ zarr + handheld_gripperur5 + + + + + + +
+
+ +
+ +
+ + + + 1,400 episodes + + + + + + + MIT + + +
+ +
+
+
+ + +
+
Paper
+ + + arXiv:2402.10329 + +
+ + + +
+
Tags
+
+ manipulationreal worldcontact rich +
+
+ + + +
+
Task Types
+
+ pick placepushcontact rich +
+
+ + +
+
Sources
+
+ +
+ http + + https://umi-gripper.github.io/ + + Example datasets released as Zarr; see project page / GitHub + +
+ +
+
+ +
+
Forge Commands
+
+
+ forge inspect umi + +
+
+ forge quality umi + +
+
+ forge convert umi ./output --format lerobot-v3 + +
+ +
+
+ + +
+
Notes
+
Human handheld-gripper collection paradigm (no robot needed to collect). Zarr-native.
+
+ + +
+
+
+
+