Reproduce “SignVerse-2M: A Two-Million-Clip Pose-Native Universe of 55+ Sign Languages” 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
Sen Fang, Hongbin Zhong, Yanxin Zhang, Liwei Che, Dimitris N. Metaxas
SignVerse-2M converts public sign-language videos into standardized DWPose sequences of body, hand, and facial keypoints, with subtitle-timed clips spanning more than 55 languages. It supports text-to-pose modeling and pose-conditioned rendering. On the same ASL split, SignDW Transformer (1.2B) outscored Progressive Transformers on BLEU-4, ROUGE, and DTW-MJE; these cross-architecture results demonstrate usability, not an isolated benefit of the representation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.