I have previously outlined my goal of testing multiple photogrammetry solutions on a single dataset, and reporting times and results.
I’m using a dataset based on photographs of this Styracosaurus model (I’ve had it since I was quite young):
The dataset has 53 photos in total, and is available from this link. [This will be moved to figshare in due course].
The model is about 12 cm in total length, has texture for the scales, and a reflective brass nameplate on the base. The model was situated on a glass desk, and there may be reflection issues from that.
Previously, I ran through sparse reconstruction with OpenMVG, followed by dense reconstruction with MVE. Today, I’ll use MVE (or Multiview Environment) to complete the whole process.
I’m running everything directly from the binaries directory for MVE, and creating a new project within there called ‘styracMVE’, which is set up with the following command:
makescene.exe -i ..\Blog\Photogrammetry\Styrac\ styracMVE
This tells the makescene executable to create a new folder called ‘styracMVE’ in the current directory, and use the images in ..\Blog\Photogrammetry\Styrac as the source.
Time taken: 12.5 seconds
The we need to reconstruct the camera locations with sfmrecon (structure from motion reconstruction):
That is, reconstruct based on the scene input we just created.
Time Taken: 27 seconds
After this, the process is identical to that used in the previous post, except that this time I’m using windows binaries instead of Linux (so the ‘/’ changes to ‘\’):
Time taken: 549 seconds
scene2pset.exe -F2 styracMVE styracMVE\pset-L2.ply
Time taken: ~10 seconds
fssrecon.exe styracMVE\pset-L2.ply styracMVE\surface-L2.ply
Time taken: 1457 seconds
meshclean.exe -t10 styracMVE\surface-L2.ply styracMVE\surface-L2-clean.ply
Time taken: 6 seconds
And here’s the uncoloured and coloured result:
Total Time: 2061.5 seconds
So, slightly quicker than using OpenMVG to match cameras, and al almost identical result.
Well, there really is very little difference with using openMVG to do the matching, at least with the basic parameters I used each time, but it was about 3-4 minutes quicker overall. The model is almost identical – perhaps a little more complete around the base, but the same areas are still missing.
Final model (downloadable from sketchfab):