Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
UMR (Unified Motion Retargeting for Humanoids with Learned Point Cloud Correspondence) is a research project for transferring human motion onto humanoid robots. It represents body surfaces as sets of 3D points and learns correspondences between the human and robot shapes, with the aim of preserving contacts during retargeting.
According to a YouTube creator summary, it handles demanding motions such as spin kicks, stair climbing and tennis, and code has reportedly been released. The project page is the only source reviewed here; details on architecture, benchmarks and supported robots were not available in the fetched material.
It is aimed at robotics researchers working on humanoid motion imitation. The vendor or authoring institution, license and repository location could not be confirmed from the provided material.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Retargets human motion to humanoid robots by representing body surfaces as 3D point sets and learning correspondences, preserving contacts. Handles spin kicks, stairs and tennis; code released.
Closest neighbours by meaning, not just by category.
Model
Monocular video motion capture that combines flow matching with physics-based RL for plausible 3D human motion.
Hardware · Ant Group
Low-cost upper-limb exoskeleton for teleoperating robot arms with real-time haptic torque feedback, from Ant Group-backed research.
Research
Research method that retargets human hand-object demonstrations to dexterous robot hands using one multi-reference RL policy.
Research
Research system that transfers real tennis players' motions onto a Unitree G1 humanoid for rallies and serves.
Research
Hierarchical humanoid loco-manipulation framework that chains meta-skills via contact flow, accepted to CoRL 2026.
Research
Research framework letting a frozen VLM agent learn robot tasks in context from videos, images and feedback, no fine-tuning.