Smartphone Video-Based Monocular 3D Reconstruction

Neural radiance fields (NeRFs) have rapidly progressed in novel view synthesis, enabling photorealistic 3D scene reconstruction. However, achieving precise 3D models requires expensive equipment. The primary focus of this study is enhancing NeRF-based 3D reconstruction using RGB smartphone sensors. In this work, we provide a guide for capturing videos of indoor and outdoor scenes to ensure successful 3D reconstruction using a smartphone. This includes algorithms for automatically removing artifacts and irrelevant points from the captured video data.