#surface reconstruction

ObjectSDF++ offers a novel approach to enhance surface reconstruction accuracy for both scenes and objects through the use of an occlusion-aware opacity rendering technique and object-distinction regularization. By incorporating instance mask supervision, it provides precise neural implicit surfaces suitable for 3D modeling and computer vision applications. The project leverages existing tools such as MonoSDF, Omnidata, and vMAP, facilitating straightforward setup and thorough evaluation with provided scripts.

Explore the Gaussian Opacity Fields (GOF) project, which utilizes 3D Gaussians and Marching Tetrahedra to streamline surface reconstruction in expansive environments. This technique enhances mesh extraction efficiency and speed, with recent advancements doubling training pace through optimized operations. Featured at SIGGRAPH ASIA 2024, GOF demonstrates exceptional performance on datasets such as TNT and Mip-NeRF 360.