Tutorial
Structure from Motion and Open Multi-View Stereo Reconstruction
3D reconstructions, derived from Structure from Motion (SfM), have broad applications in virtual reality, cultural heritage preservation, architectural modeling, robotics, and medical imaging, facilitating accurate mapping and navigation. These techniques utilize geometric principles and image-based methods to reconstruct 3D scenes from 2D images.
The objective of this tutorial session is to introduce participants to the core concepts of SfM and Multi-View Stereo (MVS) reconstruction using open-source libraries like openMVG (Multiple View Geometry) and openMVS (Multi-View Stereo). Participants will gain hands-on experience processing image datasets to create 3D reconstructions from various viewpoints. The session will cover:
Fundamental principles and significance of SfM and MVS in computer vision, robotics, and augmented reality.
Introduction to openMVG's functionalities for SfM, encompassing feature detection, matching, and camera pose estimation, alongside an overview of openMVS for dense reconstruction, meshing, and texturing.
Installation guidance for openMVG and openMVS on different platforms, including setting up essential dependencies and libraries.
Tips for capturing images suitable for successful reconstruction, covering aspects like camera calibration, image quality, and overlap.
Practical demonstrations using openMVG for SfM, including feature detection, matching, incremental Structure from Motion, and global bundle adjustment.
Progressing to openMVS for dense reconstruction, covering mesh reconstruction, refinement, and texturing.
Techniques for refining and optimizing the resultant 3D model, followed by visualization using open-source tools.
Repositories
GitHub - openMVG/openMVG: open Multiple View Geometry library. Basis for 3D computer vision and Structure from Motion
https://github.com/openMVG/openMVGGitHub - cdcseacave/openMVS: open Multi-View Stereo reconstruction library
https://github.com/cdcseacave/openMVS