humanml3d

[Open-source Project] UniMoCap: community implementation to unify the text-motion datasets (HumanML3D, KIT-ML, and BABEL) and whole-body motion dataset (Motion-X).

🎭 Enhance facial expression recognition using multimodal large language models to improve emotion analysis in various applications.

Text to motion model (pytorch) trained on HumanML3D dataset

Code for "Which Geometry Predicts Few-Step Degradation in Latent Flow Matching?" (BeNTo @ NeurIPS 2026)

Physics-informed continuous autoencoder for compact 3D human motion

Text-to-3D-human-motion in 7 languages: a controlled study extending MotionCLIP with M-CLIP, per-language encoders, a VAE objective and contrastive alignment.

Repo for portfolio, containing working redirects to all projects.