datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
SMPL_Humanoid_offline_dataset
UniPhys Dataset: Offline Dataset for Physics-based Character Control
This dataset is part of the UniPhys, enabling large-scale training of diffusion policies for physics-based humanoid control using SMPL-like characters. The state-action pairs are generated by the PULSE motion tracking policy.
Dataset Overview
amass_state-action-pairs: state-action pairs for motion sequences from AMASS dataset (excluding infeasible motions)
babel_state-action-text-pairs: Packaged… See the full description on the dataset page: https://huggingface.co/datasets/yan0116/SMPL_Humanoid_offline_dataset.RoMo-SMPL
RoMo-SMPL — In-the-Wild SMPL Body Motion (RoMo Paper Core)
RoMo-SMPL is the paper-aligned release of the RoMo body motion corpus in SMPL body parameter space (global orientation, 21-joint body pose, shape, translation). Each clip includes five text captions and a three-level semantic taxonomy (category, subcategory, atomic action), with fixed train / val / test splits.
Paper: RoMo: A Large-Scale, Richly Organized Dataset and Semantic Taxonomy for Human Motion Generation… See the full description on the dataset page: https://huggingface.co/datasets/RoMoDataset/RoMo-SMPL.ViViD_SMPLXSMPLX-amass
Dataset Card for "SMPLX-amass"
More Information needed
eyesjapan-mocap-smplx
Robot Motion Application Previews
The examples below demonstrate how motion sequences from this dataset can be retargeted and applied to robotic motion, illustrating potential uses in humanoid-robot motion generation, control, and imitation learning.
Karate — Mawashi Geri
Low Kick — Left
Example demonstrations illustrating potential applications of the motion-capture sequences provided in this dataset to robotic motion.
Eyes, JAPAN MoCap NPZ… See the full description on the dataset page: https://huggingface.co/datasets/eyesjapan/eyesjapan-mocap-smplx.SMPL_samples
SMPL motion dataset
Episodes
Episode
Motion
0
dance_hiphop
1
dance_jackson
2
eating_apple
3
high_jump
4
hiphop_ii
5
horse_riding
6
jump_360
7
mambo_kicks
8
reach_jump
9
turn_jump_270
10
turn_jump_360
11
victory_dance
12
walk_forward
SynthMoCap_smpl_512https://github.com/microsoft/SynthMoCap
Rendered the SynthMoCap dataset as smpl.
models_smplx_v1_1smplxSynthMoCap_photo_to_smpl_512https://huggingface.co/datasets/toyxyz/SynthMoCap_smpl_512
SMPL-Interact-Plussmpl_train_seed42unveil-smpl-binssmpl_train_seed8smplestsmplsmplx_anything
SMPLX Anything
smplx_anything is a unified-format preprocessed bundle (processed_*_resized) of four SMPL-X-based 3D human datasets, packaged into a single tar archive for easy distribution.
Sub-datasets
Sub-dataset
Folder
Project page
Paper
AGORA
processed_agora_resized/
agora.is.tue.mpg.de
Patel et al., CVPR 2021
Anny-One
processed_annyone_resized/
Anny-One @ NAVER LABS Europe · GitHub
Baradel et al., 2025
BEDLAM… See the full description on the dataset page: https://huggingface.co/datasets/Yong-Hoon/smplx_anything.humanml3d-smpl-genmo
GENMO-ready HumanML3D-SMPL
This private dataset contains the HumanML3D-SMPL files used for GENMO training.
Folder structure
hmr4d_support/
humanml3d_smplhpose_train.pth
humanml3d_smplhpose_val.pth
humanml3d_smplhpose_test.pth
humanml3d_smplhpose_train_val.pth
t5_embeddings_v1_half/
all_text_embed.pth
test_text_embed.pth
Use in GENMO
Place this folder at:
CausalGENMO/inputs/HumanML3D_SMPL
Example:
cd ~/SageMaker/CausalGENMO
mkdir -p inputs
ln -sfn… See the full description on the dataset page: https://huggingface.co/datasets/nahgetout/humanml3d-smpl-genmo.fame-smplx-meshsmpl-assets100style_smplxamass_smplhsmu_counsel_smplSeamless-Avatar-Smplx-150h
Seamless Avatar SMPLX 150h Dataset
📊 Dataset Composition & Splits
The dataset is curated based on a total duration of 150 hours. The distribution is managed according to the following ratios:
Split Ratio:
Train: 60% (90 hours)
Dev: 20% (30 hours)
Test: 20% (30 hours)
Interaction Types:
Naturalistic: 2/3 of the total duration.
Improvised: 1/3 of the total duration.
🦴 SMPL_X Model Construction
Our SMPL_X data represents a high-fidelity full-body… See the full description on the dataset page: https://huggingface.co/datasets/xwshi/Seamless-Avatar-Smplx-150h.humanml3d_smplxsmu_counsel_smplsmu_counsel_smplsmu_counsel_smplsmu_counsel_smplsmu_counsel_smpl