CMVS/PMVS2 40% faster

Test data fir Nghia Ho implementation

We tested the implementation written by Nghia Ho and the results are amazing. Data elaboration was effectvively 40% faster. Now we are working on the new package for ArcheOS, wich will be released ASAP.
In the image you can see the data we used for the test: the old version of CMVS/PMVS needed 153 minutes, while the new implementation completed the process in just 92 minutes.

Good news from PMVS2

According to Pierre Moulon's Blog a new implementation (from Nghia Ho) has been integrated in CMVS-PMVS. The result should be a faster (up to 30%) data elaboration. We will start with the test as soon as possible and, in case, a new ArcheOS package for PPT (Python Photogrammetry Toolbox) and PPT-GUI will be released. For who wants to collaborate in PPT-GUI development, here is the link on GitHub.

In the video below are some archaeological examples done with this software.

Python Photogrammetry Toolbox (PPT) and a turntable for 3D objects

I already answered on Pierre Moulon' Blog, but the question Tauricity did is very interesting. The question regards the possibility to document with PPT (Python Photogrammetry Toolbox) an object on a turntable, keeping the camera on a fix position. I had positive results when I tried with Giuseppe Naponiello (Arc-Team's anthropologitst) to document an human skull. We put the skull on a turntable with a black board behind (in this way the software does not recognize nothing exept the turntable and the skull); then we took a picture every 30 degrees. The only problem we had was during data elaboration, cause using PPT with Bundler 4 the software gave us a wrong result (it seems that it confused the two faces of the skull in the axial simmetry). Alessandro Bezzi helped us elaborating the data via shell and using Bundler 3 instead of Bundler 4 (cause he had a similar problem documenting the Theodoric Mausoleum in Ravenna) and this time there were no errors and we could do a dense pointcloud inside PMVS2. So it looks like that Bundler 4 has problems when objects have an axial (the skull) or a radial (the mausoleum) simmetry, but bundler 3 works perfectly.

Anyway it was possible for us to document an object on a turntable keeping the camera fix; the only tip was the black board to "erase" the background, so the software does not recognize points in it (which would be in conflict with the points of the object).

In this video is possible to see the result of the dense pointcloud.