Main content
Top content
Evaluation of multi-view 3D reconstruction software - CAIP2015
A number of software solutions for reconstructing 3D models from multi-view image sets have been released in recent years. Based on an unordered collection of photographs, most of these solutions extract 3D models using structure-from-motion (SFM) algoritdms. In this work, we compare the resulting 3D models qualitatively and quantitatively. To achieve these objectives, we have developed different metdods of comparison for all software solutions. We discuss the perfomance and existing drawbacks. Particular attention is paid to the ability to create printable 3D models or 3D models usable for other applications.
On this page we present additional material regarding our paper Evaluation of multi-view 3D reconstruction software [1] in the proceedings of the 16th International Conference on Computer Analysis of Images and Patterns (CAIP'15). To view the 3D point clouds please click on the corresponding image.
Additional Material (click on the images)
Data set | PhotoScan | 123D Catch | VisualSFM | ARC 3D | Ground Truth | |
P11 | ||||||
P8 | ||||||
Oxford Dinosaur | N/A | |||||
Oxford Dinosaur* | N/A |
Created 3D models of PhotoScan, 123D Catch, VisualSFM and ARC 3D on four data sets: fountain-P11 (11 images), Herz-Jesu-P8 (8 images), Oxford Dinosaur (36 images) and *Oxford Dinosaur with removed background (36 images). One sample image of each data set is shown on the left, reconstruction results of each tool are shown in the center and the corresponding ground truths are shown on the right.
Heat Maps (click on the images)
(a) | (b) |
(c) | (d) |
Heat map of the minimal distance between the ground truth point cloud and the triangular mesh of the created model as reference on the fountain-P11 data set. Points with distance differences more than ±0.1 are not visualized; distance differences between -0,05 and 0.05 are colorized in the scheme blue-green-red. On the right next to the legend, the distance distribution is shown. Models are created by (a) PhotoScan, (b) 123D Catch, (c) VisualSFM and (d) ARC 3D.
(a) | (b) |
(c) | (d) |
Heat map of the minimal distance between the ground truth point cloud and the triangular mesh of the created model as reference on the Herz-Jesu-P8 data set. Points with distance differences more than ±0.1 are not visualized; distance differences between -0,05 and 0.05 are colorized in the scheme blue-green-red. On the right next to the legend, the distance distribution is shown. Models are created by (a) PhotoScan, (b) 123D Catch, (c) VisualSFM and (d) ARC 3D.
(a) | (b) | (c) |
(d) | (e) | (f) |
Heat map of the minimal distance between the ground truth point cloud and the triangular mesh of the created model as reference on the original Oxford Dinosaur data set (a)-(c) and on Oxford Dinosaur with removed background (d)-(f). Points with distance differences more than ±0.01 are not visualized; distance differences between -0.005 and 0.005 are colorized in the scheme blue-green-red. On the right next to the legend, the distance distribution is shown. Models are created by (a),(d) PhotoScan, (b),(e) 123D Catch and (c),(f) VisualSFM.
References
[1] | J. Schöning & G. Heidemann. Evaluation of Multi-view 3D Reconstruction Software.<br/>In Computer Analysis of Images and Patterns, 9257 : 450-461, ISBN: 978-3-319-23116-7, 2015. Springer International Publishing. | PDF | DOI | URL | BibTeX |