← Sinan Lab › All datasets › MimicGen Datasets
Dataset · NVIDIA (NVlabs)

MimicGen Datasets

Landing page →Hugging Face →

Simulation (robosuite/MuJoCo; Isaac Gym Factory for some tasks). Paper reports 'over 50K demonstrations across 18 tasks' from ~200 human demos; hub release is 48,000+ across 12 tasks. Modalities not enumerated on fetched pages.

Facts

Organisation
NVIDIA (NVlabs)
Form tags
single arm · simulation
Embodiments (as stated)
Panda · Sawyer · IIWA · UR5e
Episodes
48,000
Hours
unverified
Frames
unverified
Tasks
12
Scenes
unverified
Modalities
unverified
Format
HDF5 (robomimic)
Licence
CC-BY 4.0
Commercial use
allowed (stated)
Released
2023-10
Size
149 GB
Sources · 7 sources
embodiments https://mimicgen.github.io/ Official page · Fetched 2026-09-18 · No snapshot stored yet 'Panda', 'Sawyer', 'IIWA', 'UR5e' named as robot arms in the 'robot' dataset group
episodes https://github.com/NVlabs/mimicgen Official page · Fetched 2026-09-18 · No snapshot stored yet 'The released datasets contain over 48,000 task demonstrations across 12 tasks'; HF card breakdown: source 120, core 26,000, object 2,000, robot 16,000, large_interpolation 6,000
tasks https://github.com/NVlabs/mimicgen Official page · Fetched 2026-09-18 · No snapshot stored yet 12 tasks in released datasets (paper abstract: 18 tasks)
format https://huggingface.co/datasets/amandlek/mimicgen_datasets Official page · Fetched 2026-09-18 · No snapshot stored yet HDF5 files compatible with robomimic
license https://github.com/NVlabs/mimicgen Official page · Fetched 2026-09-18 · No snapshot stored yet 'the datasets are released under CC-BY 4.0'; HF license tag cc-by-4.0
release_date https://arxiv.org/abs/2310.17596 Official page · Fetched 2026-09-18 · No snapshot stored yet Submitted 26 Oct 2023; CoRL 2023
size_gb https://huggingface.co/datasets/amandlek/mimicgen_datasets Official page · Fetched 2026-09-18 · No snapshot stored yet 'Total file size: 149 GB'

Related bodies in the index

Matched by stated robot names or form; a match means the form is covered, not that a model was trained on it.

SinanLab · Sinan LabClear direction, grounded judgement.