TOP

3D Reconstruction with COLMAP

COLMAP could be used as an alternative for the VisualSFM & CMPMVS process. Read on for usage and performance evaluations.

 



Image Capturing


(Same as in reconstruction workflow)

Recommended equipment:
Digital camera is recommended. We used Samsung Galaxy camera, which captures images that are 4608 pixels wide and 3456 pixels high; and videos that are 1920x1080. The idea is to get better quality images. Higher resolution images usually help preserve detail and get better reconstruction result.

The images do not have to be taken with a single camera (i.e. differences on resolutions, or image sizes, are acceptable).

 



Software Installation


Software and hardware requirement:

COLMAP runs on all major platforms (Linux, Mac, Windows). On Linux, CUDA is recommended.

For installation, please refer to official documentation on COLMAP 3.3 website.

 



Performance Comparison


COLMAP vs CMPMVS:

COLMAP
CMPMVS
OS
Windows, Linux, Mac
Windows only
License
GNU General Public License
Research Only
Time
Less Slow*
Slow
GUI
Yes
No
Input
2D Digital Images
Requires VisualSFM output

Note: * COLMAP takes less time generating colored mesh reconstruction. In our example below, 297 images are used. COLMAP completes the process in 15 hours; whereas CMPMVS takes over 20 hours.

COLMAP pipeline is more comprehensive, since it takes image input and generates sparse/dense/mesh results. As shown in the mesh reconstruction results from COLMAP and CMPMVS below, COLMAP is better at smoothing out featureless surfaces.

mve_vs_cmpmvs

mve_vs_cmpmvs2

mvs_vs_cmpmvs3

Two more sets of data are used for testing. The first set contains 136 images of the exterior and interior of a Jeep. Both COLMAP and CMPMVS can successfully get a complete model. The second set are images covering the left front wheel of the Jeep from different angles. Both COLMAP and CMPMVS succeed in preserving the key details. The results and comparison are shown in the videos below.

COLMAP models are available for downloading here.

COLMAP CMPMVS
   
COLMAP CMPMVS