Reproduce “SIZER: A Dataset and Model for Parsing 3D Clothing and Learning Size Sensitive 3D Clothing” 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
Garvita Tiwari, Bharat Lal Bhatnagar, Tony Tung, Gerard Pons-Moll
Researchers at MPI for Informatics and Facebook Reality Labs introduce SIZER, the first 3D dataset capturing real-world clothing size variations on diverse subjects, and propose two data-driven models. ParserNet automatically extracts multi-layer garment meshes and underlying body shapes from single 3D scans, while SizerNet predicts size-sensitive 3D clothing deformation, enabling realistic virtual try-on.
Key insights
- A localized, sparse weighted sum approach in ParserNet effectively preserves fine geometric details and handles complex deformations during 3D garment parsing, outperforming direct displacement prediction.
- Learning explicit non-linear relationships between garment shape, body shape, and garment size from real-world data (SIZER) is crucial for accurate and realistic size-sensitive clothing prediction.
- The combination of automated parsing (ParserNet) and size-sensitive modeling (SizerNet) creates a streamlined pipeline for editing garment sizes on 3D meshes, enabling practical applications like virtual try-on.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.