Reproduce “SRM-Hair: Single Image Head Mesh Reconstruction via 3D Morphable Hair” 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
Zidu Wang, Jiankuo Zhao, Miao Xu, Xiangyu Zhu, Zhen Lei
Researchers at CASIA introduce SRM-Hair, a method that reconstructs an independent 3D hair mesh and full head from a single image. The approach leverages a novel 3D Morphable Hair model built on semantic-consistent ray modeling, demonstrating superior accuracy in capturing complex hair geometry.
Key insights
- Semantic-consistent ray modeling effectively establishes vertex-level correspondence for complex, non-rigid structures like hair, enabling the creation of a controllable morphable model similar to facial 3DMMs.
- Capturing both the nearest (`D_min`) and farthest (`D_max`) ray intersection points from the scalp provides comprehensive information to represent the full volumetric extent and depth of hair geometry.
- A refinement module significantly improves the quality of reconstructed hair meshes by identifying and excluding outlier points generated by the morphable model, addressing the inherent complexity of hair surfaces.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.