핵심 개념
Efficient and accurate scene reconstruction using a novel tensorial inverse rendering approach.
초록
Introduces TensoIR, a novel inverse rendering approach based on tensor factorization and neural fields.
Proposes a method for estimating scene geometry, surface reflectance, and environment illumination from multi-view images under unknown lighting conditions.
Demonstrates superior results in radiance field reconstruction and physically-based model estimation for realistic view synthesis and relighting.
Utilizes a low-rank tensor representation for fast and compact reconstruction while efficiently modeling secondary shading effects.
Outperforms baseline methods qualitatively and quantitatively on challenging scenes.
통계
"Our approach can accurately model secondary shading effects (like shadows and indirect lighting) and generally support input images captured under single or multiple unknown lighting conditions."
"Our method can accurately model secondary shading effects (like shadows and indirect lighting) and generally support input images captured under single or multiple unknown lighting conditions."
인용구
"Our approach can accurately model secondary shading effects (like shadows and indirect lighting) and generally support input images captured under single or multiple unknown lighting conditions."