HairLRM: Strand-based Hair Modeling via Large Reconstruction Models

2026-06-13

AI Summary

Existing hair strand–based 3D hair reconstruction methods often suffer from over-smoothing and geometric distortions due to the lack of global structural constraints, particularly under occlusion and ambiguous local orientation cues. This work introduces large reconstruction models (LRMs) into strand-level hair modeling for the first time, leveraging their mesh outputs as structural priors. We propose a dual-orientation autoencoder that refines coarse geometry into high-fidelity hair strands and integrate a surface-guided latent space optimization strategy to effectively resolve vector field singularities and disentangle complex topologies. Our approach significantly enhances reconstruction accuracy and robustness, excelling on challenging hairstyles such as ponytails and curly hair while avoiding over-smoothing, thereby establishing a new state-of-the-art benchmark for 3D hair reconstruction.

Abstract

The fundamental limitation of traditional strand-based modeling is not simply data scarcity, but the ill-posedness of inferring complex 3D fields from 2D imagery without structural constraints. This unconstrained regression leads to catastrophic failures in resolving both global occlusion (e.g., in ponytails) and local directionality (e.g., in curls), resulting in over-smoothed, plausible-but-incorrect geometries. To resolve this, we integrate the strong geometric priors of Large Reconstruction Models (LRMs) into the strand generation pipeline. Using the LRM mesh as a structural anchor, we employ a novel Dual Orientation AutoEncoder to lift coarse geometry into high-fidelity strands. By resolving vector field singularities through latent-space optimization and surface-guided refinement, our method effectively disentangles complex topological structures, setting a new benchmark for robustness and accuracy in hair reconstruction.

Problem

• strand-based hair modeling
• ill-posed reconstruction
• 3D hair geometry
• occlusion handling
• directional ambiguity

Innovation

• strand-based hair modeling
• Large Reconstruction Models
• Dual Orientation AutoEncoder
• vector field singularities
• surface-guided refinement