Tamaththul3D: High-Fidelity 3D Saudi Sign Language Avatars from Monocular Video
A new reconstruction pipeline named Tamaththul3D has been developed by researchers to create high-fidelity 3D avatars of Saudi Sign Language (SSL) from monocular video input. This innovation fills a significant void in Arabic Sign Language (ArSL) technology, which caters to around 400 million Arabic speakers globally but lacks effective 3D reconstruction methods and high-quality parametric annotations. The team has produced the first high-quality 3D parametric annotations for the Ishara-500 SSL dataset, detailing accurate SMPL-X parameters for 500 culturally relevant SSL signs. Tamaththul3D employs SMPLer-X for body estimation, WiLoR for enhanced hand refinement, and MediaPipe for 2D pose oversight, achieving top-tier hand accuracy through advanced techniques. This advancement is crucial for enhancing the visibility and accessibility of Saudi Sign Language within digital environments.
Key facts
- Tamaththul3D is a reconstruction pipeline for 3D SSL avatars from monocular video.
- It provides the first high-quality 3D parametric annotations for the Ishara-500 SSL dataset.
- The dataset includes precise SMPL-X parameters for 500 culturally authentic SSL signs.
- The pipeline integrates SMPLer-X, WiLoR, and MediaPipe.
- It uses kinematic-chain-based wrist alignment with hybrid swing-twist decomposition.
- The work addresses a gap for 400 million Arabic speakers.
- The research is published on arXiv with ID 2605.05367.
- The pipeline achieves state-of-the-art hand accuracy.
Entities
Institutions
- arXiv
Locations
- Saudi Arabia