Reproduce “ETA-Motion: Efficient Topology-Agnostic Motion Diffusion via Sparse Joint Learning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Seoyoung Hwang, Tae-Kyun Kim
ETA-Motion generates motion for characters with different skeletons by training on a compact set of representative joints and using the target skeleton’s full rest pose at inference. It models joint positions and supports varying joint counts. In a 300K-step ablation, K=20 achieved 65.73% Unseen Topology Coverage versus 42.45% for full-joint training; mesh animation still requires inverse kinematics.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.