Reproduce “MoPort: A View-Aware 2D Interface for Motion Language Models” 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
Lixing Tan, Qing Xia, Yuting Guo, Shuai Li, Aimin Hao
MoPort maps 2D joint tracks from monocular video into a pretrained motion-language model’s native tokens without reconstructing 3D motion. It combines paired-view alignment with camera geometry to address viewpoint changes. On HumanML3D, it reaches R@1=0.5674, versus 0.4537 for 2D-Interface. This result covers sampled camera views; the method assumes known intrinsics and still fails on some out-of-distribution views and motions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.