AnyMo: Geometry-Aware Setup-Agnostic Modeling of Human Motion in the Wild

Researchers introduced AnyMo, a geometry-aware model for human motion understanding from wearable IMU data.

AnyMo combines physics-based IMU simulation, graph pretraining, motion tokenization, and language alignment to generalize across unseen sensor placements.

The model improves zero-shot recognition, cross-modal retrieval, and motion captioning on multiple datasets.

The work addresses a key portability challenge in wearable sensing by making motion models more transferable across devices and setups.