NeuralMeshing: Complete Object Mesh Extraction from Casual Captures

TL;DR

NeuralMeshing is an automated system that reconstructs complete 3D object meshes from multiple casual videos by combining structure-from-motion with known point references, eliminating the need for commercial scanners.

Contribution

It introduces a novel method for complete object mesh extraction from casual videos using known points and multi-view merging, without relying on hole filling.

Findings

Successfully reconstructs complete object meshes from casual videos

Uses known points for automatic frame alignment

Eliminates the need for hole filling in mesh generation

Abstract

How can we extract complete geometric models of objects that we encounter in our daily life, without having access to commercial 3D scanners? In this paper we present an automated system for generating geometric models of objects from two or more videos. Our system requires the specification of one known point in at least one frame of each video, which can be automatically determined using a fiducial marker such as a checkerboard or Augmented Reality (AR) marker. The remaining frames are automatically positioned in world space by using Structure-from-Motion techniques. By using multiple videos and merging results, a complete object mesh can be generated, without having to rely on hole filling. Code for our system is available from https://github.com/FlorisE/NeuralMeshing.