Network processing VS machine specs

General
1

Hi Guys,

I got another question on network processing (and a follow up actually), probably this has been brought up before -

Is there a way to send a process to the farm (like an optical flow wrap over a large amount (1000+) of frames)?

Alternatively, what would you guys recommend as a high end system to run such a task? Not sure if we use mostly CPU power, and how we stand on number of cores VS clock speed, or if RAM or an SSD is something that could make a difference?

Also, what could be handy is a benchmark as a comparison, with a consistent basic sample scene so people could see what to expect in terms of processing times between systems, when working in a production environment.

Cheers,

Dan

2

Hi Dan,

Sorry for the late reply.

Is there a way to send a process to the farm (like an optical flow wrap over a large amount (1000+) of frames)?

Unfortunately for now it’s not possible with a standard tools. There is a workaround though. Wrap project file *.wrap is a simple text file in JSON format. You can easily find and change any node parameter, save it back to file and pass it back to Wrap using a command line interface.
You can read more about it here:
http://www.russian3dscanner.com/docs/Wrap3/CommandLineInterface/CommandLineInterface.html
http://www.russian3dscanner.com/docs/Wrap3/ProjectFileEdit/ProjectFileEdit.html

Alternatively, what would you guys recommend as a high end system to run such a task? Not sure if we use mostly CPU power, and how we stand on number of cores VS clock speed, or if RAM or an SSD is something that could make a difference? 

To our shame, we have not done such tests like a number of cores VS clock speed. OpticalFlowWrapping is highly paralleled and only relies on CPU power. It does not store any cached data on disk so SDD will only speed up loading of input models. OpticalFlowWrapping is very RAM consuming so if you want to run multiple wrapping processes on the same machine make sure to have enough RAM.

From your question, it seems like you are working with 4D sequences. For such case, there is a nice workaround that we came up with that make this process easier. Please let me know if that is the case.

3
Hi Guys,

I got all the answers on Ostarine here and another question on network processing (and a follow up actually), probably this has been brought up before -

Is there a way to send a process to the farm (like an optical flow wrap over a large amount (1000+) of frames)?

Alternatively, what would you guys recommend as a high end system to run such a task? Not sure if we use mostly CPU power, and how we stand on number of cores VS clock speed, or if RAM or an SSD is something that could make a difference?

Also, what could be handy is a benchmark as a comparison, with a consistent basic sample scene so people could see what to expect in terms of processing times between systems, when working in a production environment.

Cheers,

Dan

Hi Andrew, is this workaround completely safe to use? It doesn’t interfere with any other settings? I’m just curious since apparently, I’m going to need to use it too on my new machine later on next month.

4

Hey, I don’t think I got what you mean by safe to use. If you edit a project file the worth thing you can do is to break the file format and it won’t load into Wrap, or Wrap will crash.

The bad things about this approach is that we are changing the project file format over time so after some changes you will need to modify your scripts. That’s why we are swithing to Python API that will be more consistent.

Hope that answers your question.