The author introduces UFORecon, a robust view-combination generalizable surface reconstruction framework that outperforms existing methods in both favorable and unfavorable scenarios.
提案されたUFOReconフレームワークは、交差ビューマッチング機能を活用して画像間の相関を学習し、明示的な2Dマッチング類似性をエンコードすることで、従来の表面再構築手法よりも優れたパフォーマンスを実現します。
A novel hybrid directional parameterization that effectively combines the advantages of viewing and reflection directions to handle both specular surfaces and complex structures in neural implicit surface reconstruction.
AniSDF presents a unified approach that learns fused-granularity neural surfaces with anisotropic encoding to achieve high-quality 3D reconstruction and photorealistic rendering.