venerdì 24 marzo 2017

Making snap packages of photogrammetry software

Some time ago I got vaguely interested into photogrammetry, that is the reconstruction of a 3D model out of a set of plain 2D photographs. I just thought that it was cool, and wanted to try it.

Unfortunately, the most popular of these tools, VisualSFM, was not packaged for Ubuntu and didn't come with ready binaries. Furthermore, the steps to build it are far from trivial: they include modifying a few of the source files!

So, while I was going through this ordeal in order to compile it, I though of how I could avoid running through all this pain once more, had the need emerged to build this program again in the future. I initially thought of writing a shell script to automate it, but then I realized that there exists a much better solution: a snapcraft recipe!
This solution has the big advantage that the resulting binary (called a "snap" package) can be shared with other Linux users, by publishing it into the snap store. Therefore, one doesn't need to be a programmer or a computer expert anymore in order to install the software.

As I quickly found out, other "structure from motion" and "multi-view stereo" (the two parts of the 3D reconstruction pipeline) programs are also unavailable as binaries for Linux, and require quite some effort to be built. As a matter of fact, this problem is quite common for scientific and academic software: always written by authentic geniuses in the field of research, but who often are not as experienced (or interested) in software distribution.

So I thought — well, given that I've just made a snap package (and that I've even enjoyed the process!), why stop here? :-)

And here you have it: most of this photogrammetry software is now available as snap packages, which makes it trivial to install them and try them out. Though indeed, the 3d reconstruction can take a lot of time, so that's another thing to be considered.

To help you out in deciding which software to use, I made a video review of structre from motion and multi-view stereo tools; without any pretense of scientificity, — be it clear! — just with the goal of giving an overview of what is available out there, and how easy (or difficult!) to use it is:

I also "snapped" a couple of other programs related to 3D reconstructions. One of them is CloudCompare, a 3D point cloud and mesh processing software.

I might be making more videos on this subject, so stay tuned. :-)

Etichette: , , , , , ,

4 Commenti:

Blogger WEBmadman ha detto...

Thank you so much for the tutorial and the Snaps as well. I had been using the Python Photogrammetry Toolbox, but when I moved from Ubuntu 14.04 to 16.04, it stopped working. :-(

After spending too many hours down curse filled rabbit holes, I found this- Yay!

I'm finding the best results from the MVE workflow.

I have been also working with Blend4web, you can find my first experiment with the new workflow in this folder-

The html file s the final result, the video is where the images for the reconstruction came from and the folder contains the actual images.

Basically, I'm working on a workflow that can quickly capture and create 3d models from the "real" world and present them on the web- and you have helped me achieve that- thank you!

19 aprile 2017 00:55  
Anonymous Anonimo ha detto...

The video is great, and snap packages is a huge pain saver ! Big thanks !

10 maggio 2017 13:32  
Blogger Gregorio Martinez ha detto...

Hello Mardy i have some questions, first the script is not making the file bundle.out, i get other two files but not this one, however to get the point cloud in meshLab i used the file named bundle.init.out, using the same set of pictures as you use in the video i am getting less than 600 points in the cloud, my question is, is there a way to get more points forcing the script, or how to generate the bundle.out file which perhaps has more points


31 maggio 2017 23:15  
Blogger Fabrice Kerzerho ha detto...

Hi, many thanks for those snap packages and quick tutorial which will save me a lot of time; I'm gonna test MVE now ...
Best regards,

27 agosto 2017 11:23  

Posta un commento

Iscriviti a Commenti sul post [Atom]

<< Home page