Same settings - different results

Development Asset Creation
,
1

Hello,

 

I had project with cca 850 photos. Image downsample for creating depth maps and detail decimation factor was set to 2, default grouping factor is 1.

Result was mesh with cca 45M triangles. Since original photos was not enough, I went to field again and took extra cca 100 photos, re-run alignment and reconstruction. I’ve noticed that second reconstruction is much slower than first one. Result of the second reconstruction is model with 651M triangles. First one took about 3-4h, while second one last more than 17 hours (17:32:03 to be exact), and on the end I’ve got message “Computatation of one or more parts failed”. All settings were the same. Whats the catch?

 

Thanks

 

2

3

Hi Nino,

Thank you for reporting the problem. The message means that some parts of the final model have not been computed. This is a known issue and we are working on the solution.

Here are some steps you can try:

  1. Clear the cache ( WORKFLOW \ Settings \ Clear cache \ Clear )
  2. Always use a reconstruction region.
  3. Change the RECONSTRUCTION REGION slightly, say +10-15 % bigger and rotate it by say 10-15 degrees. then reconstruct it again…
  4. If you are using the reconstruction region and the problem is still there, you can try to reconstruct the scene in blocks (https://support.capturingreality.com/hc/en-us/articles/115001483771). This is rather a workaround than the solution, but it can help deliver results in a limited time in certain cases.

If it is not a lot of data (and if it is not a problem for you), then it would be helpful to have the dataset including the project file and project folder (before the model computation) and a README.txt file with detailed steps to reproduce the problem. It is always good to have the problematic data just for internal testing.

4

  1. Next time

  2. I am :slight_smile:

  3. It appears that nothing important wasn’t computed (luckily) :slight_smile:

  4. Will do if problem appears again

 

Acutally it is, and im in the field office withoud decent internet connection :confused:

 

Thanks

 

 

 

 

5

One more question that was not answered :slight_smile:

 

Why did I get different results with the same settings (15x “heavier” model)? Is this bug related to cache or what?

 

Thanks

6

https://support.capturingreality.com/hc/en-us/articles/115001483931-How-long-is-the-computation-time-

7

Hmmm, I don’t see answer in that article :smiley:

 

So, first time I had 850 photos and computation was 3-4hrs and 45M triangles, than I added extra 100 photos (size of the scene was not changed) and compututation was 17:30hrs and 650M triangles :o

8

The last sentence from there may be the reason: “Total time of processing depends on quantity and quality of data inputs, as well as their mutual overlap, and your hardware performance.”

So there were some extra images whose number and mainly quality probably made it more difficult for the program to process it. There may also be some other processes running at the computer running draining your RAM, GPU… This can also occur when you clear the cache, which forces the application to compute some data anew.

9

Hi Nino,

inconsistencies like this have been reported before. Not sure what caused it but maybe you’ll find something if you look in the forum. I’m pretty certain that a jump from 3-4 to 17 hours cannot be explained by 100 images more. Also not the jump in Tris, which is enormous. My bet would be on some project file corruption or problem with the images. What happens if you start everything from scratch (in a new project file with empty chache)?

10

There is this strange bug where process is stuck at certain percentage. You need to pause the process and after few minutes unpause and after that everything works fine again :slight_smile: