Not enough video memory

Development Asset Creation
,
1

 

We are using a HPZ1G3 with the following specs:

Xeon Processor

32 GB RAM

1 TB SSD

NVIDIA Quadro M2000M (4GB memory)

 

We are running a project that consist of about 150 pictures, 22MB (average) each file, (.CR2).

The “aligning” process goes well and fairly quickly (It takes about 5 minutes to align).

When doing “reconstruction” in “high detail” mode it takes about 3 hours and 40 minutes to process, but at the of this process a message that says “not enough video memory”.

When doing “reconstruction” in “normal mode” it takes about 42 minutes to process, without getting any error messages.

The point is we need to work using “high detail” and generate the mesh. I read about using the clipping box tool to overcome this, but some of my colleagues do not agree with that.

Will using the clipping box tool affect the quality/size if the project?

How do I use the clipping box?

Do we need a different, more powerful video card?

Is the 40M Tris limit valid for all/any video card regardless of their capacity/performance power?

The not enough video memory error says that I’m trying to display more than 31 M tris.

I’m very new to using this software, so I hope this makes sense.

Any ideas, suggestions and feedback is greatly appreciated.

2

Hi gsappIT,

The limitation you are experiencing is a limit set by the internal viewer of RealityCapture. This arbitrary limit of 40M has nothing to do with the actual project but is only a limitation of the viewer to keep the performance up.

The clipping box you are referring to is only a visual clipping box and does *not* affect the project what so ever. The other type of box (reconstruction region) does have an impact on your project as this box tells RC which region to process into a densified point cloud and the consecutive mesh. Use this reconstruction region wisely and make sure it encompasses your entire project (tip; check your model from a side view to make sure your reconstruction region does not crop anything from the top or the bottom).

So in short. If you generate your model on “High Detail” with a proper reconstruction box you can end up with many hundreds of million (or billions) of polygons. All these polygons will be generated BUT they will not be shown in the RC viewer due to performance issues. If you use the clipping box, you can define a region to visualize the full polygons within this region. If this region contains more than 40 million triangles, then you’ll have to reduce the size of the clipping box.

If you wish to view your entire mesh model inside the RC viewer then it’s wise to decimate the polygons to 39 million. Then it will be loaded in its entirety. Otherwise you will have to export your undecimated model and open it in an external viewer.

3

Jasper, 

Thank you for your answer. This really helps.

Are you a regular user or part of Reality Capture Support ?

 

GSAPPIT

 

4

Merely an enthousiast. :slight_smile:

5

And doesn’t it show! Very clear and detailed explanation, coming from (probably painful) user experience. Includes the wider consequences of actions mentioned. Topics with posts like this I save and file as text - invaluable. Contrast with the Help files, and too much of the Support answers, written, looks like, by developers or tech people, who assume we know terms and understand things which to them are commonplace. Answers can’t be too detailed, ‘for dummies’, to exponentially increase the usefulness of the contributor’s effort in writing anything at all.

6

Hi Tom, thank you kindly. I think it’s important to help all types of users with a casual explanation. It’d be weird to assume that everyone reading these forums understands every terminology.
More users who know what they’re doing means more happy users, and thus more income for RC which means more features in future updates. :slight_smile:
Win-win-win!