3D reconstruction methods in industrial settings: a comparative study for Colmap, NeRF and 3D Gaussian Splatting

Zeno Sambugaro1, Lorenzo Orlandi1, Nicola Conci1,
1DISI University of Trento
Paper Data

Abstract

3D rendering techniques have undergone a rapid evolution with the emergence of novel and advanced methodologies, redefining the boundaries of realism and computational efficiency. This study explores recent advancements in the field, comparing established approaches like photogrammetry with software such as Colmap against the new frontiers opened by emerging so-called view synthesis approaches like Neural Radiance Fields (NeRF), and Gaussian Splatting. In this paper, we present a comprehensive comparison of aforementioned methods tailored for industrial applications, where the data acquisition campaign is generally conducted in the wild by human operators employing handheld devices.

Supplementary Results

Reconstruction on the playgrounds dataset

We show a comparison on the meshes obtained by the three compared methods:

mesh_comparison_playgroud.

We show a comparison on the neural rendering results by NeRF and 3D Gaussian Splatting, we further present a cloud-to-cloud distance comparison of the two Radiance fields based techniques with respect to photogrammetry.

pcd_comparison_playgroud_1.
pcd_comparison_playgroud_2.

Reconstruction on the excavation sites dataset

We show a comparison on the meshes obtained by the three compared methods:

mesh_comparison_excavation.

We show a comparison on the neural rendering results by NeRF and 3D Gaussian Splatting, we further present a cloud-to-cloud distance comparison of the two Radiance fields based techniques with respect to photogrammetry.

pcd_comparison_excavation.

BibTeX

@inproceedings (https://t.me/inproceedings){sambugaro20243d,
      title={3D reconstruction methods in industrial settings: a comparative study for COLMAP, NeRF and 3D Gaussian Splatting},
      author={Sambugaro, Z and Orlandi, L and Conci, N and others},
      booktitle={CEUR WORKSHOP PROCEEDINGS},
      volume={3762},
      pages={212--217},
      year={2024},
      organization={CEUR-WS}
    }