Open Source Photogrammetry with Meshroom


(Sim) #21

@michal_zurawski Thank you for the interesting paper.

For details, please read P. Moulon, P. Monasse and R. Marlet. Adaptive Structure from Motion with a contrario model estimation. ACCV 2012. You can also take a look at openMVG publications, as many parts from Alicevision are based on openMVG.

I’m asking about possibility to use targets, scale bars etc.

Meshroom is a GUI for Alicevision. Only the “core” features are implemented in the current Meshroom 2018.1.0 release. More will be implemented over time.

Alicevision supports e.g. CCTag markers. "Such marker system can deliver sub-pixel precision while being largely robust to challenging shooting conditions. github com /alicevision/CCTag. However this is not implemented in the current Meshroom 2018.1.0 release.

For details on Uncertainty Estimation read the project report (p.13-17)
Here are results from a tanksandtemples benchmark (p.10)

…can you mark a point on one photo and check if this point is automatically placed on the rest of the photos correctly?

No, not in this release

(Michal Zurawski) #22

It is from 19 to 22 but ok, i got it. It is not what i meant. I understand principle but what is important is that an information added to the model has an error in scale. So if i used CCTag, i know the distance between two of them, software will calculate an error in units of the scale or percentage? This information is very important for archivising scans.

Is there any calendar or list of features that will arrive in futre realeases? Maybe there is option to ask dev’crew for some features first, of course with proper explanation why.

Lovely. This is what make software, for me, photogrammetric.

Ok, so at this moment, GUI does not allow to obtain the photogrammetric results but it may change in future release. Thanks for the reply,

By the way, I read the paragraf about 360 cameras. Is there any chance for support in Meshroom? The problem is that data from 360 cameras, before stiching, are written sometimes to one bitmap. As they are not separate it is quite dodgy tu separate them manually and use as single photos. In my opinion, 360 are insanely interesting thing for SFM. I did some tests and i see huge potential. That is why this issue is interesting for me too.