Great, thanks Sandy and Stefan.
I will update, and build a VTK based mapper.
Thanks for your advice.
M
On 8 May 2013, at 15:36, "Engelhardt, Sandy"
<[email protected]<mailto:[email protected]>>
wrote:
Hi Matt,
the branch has been merged before ITK4 integration, so you can set your master
to:
http://mitk.org/git/?p=MITK.git;a=commit;h=1ec24168bc147499896d8db898bba4b31caec00f
Best,
Sandy
-----Ursprüngliche Nachricht-----
Von: Clarkson, Matt [mailto:[email protected]]
Gesendet: Mittwoch, 8. Mai 2013 16:25
An: Kislinskiy, Stefan
Cc: Engelhardt, Sandy; mitk-users
Betreff: Re: Bug 14890: Very Slow Rendering of PointSets
Hi Stefan,
thanks. I also thought about something that would render points, with an
associated colour value.
Imagine a data set, eg. mitkPointCloud which contains a std::vector< std::pair
< mitk::Point3D, itk::RGBPixel < unsigned char > > > (for each timestep
obviously).
Then each point can have a colour... or provide overloaded functions that
enable you to set a point without a colour, and the colour defaults to white.
Then a mapper can either render according to a set colour, provided as a colour
property on the data node.
Or it can render a point with the colour stored with the point.
And the rendering is litterally just one open GL point.. so no VTK at all?
Would this be of any use? Would it be better/worse?
M
On 8 May 2013, at 15:09, "Kislinskiy, Stefan" <[email protected]>
wrote:
Hi Matt,
I suggest you to write your own "FastPointSetVtkMapper3D". It should be quite
easy to do so. Here are some tricks to really speed things up:
1. Create only one vtkPoints, vtkCellArray, vtkPolyData, vtkPolyDataMapper,
vtkActor and don't do it in every frame. That's a one shot initialization.
2. Prefer the SetData() method of vtkPoints and hence create a vtkFloatArray
(or vtkDoubleArray) during initialization as well. You can tell the
vtkFloatArray to use your point-array without copying it.
Pseudo-code:
vector<Point> points; // Your point list
vtkFloatArray->SetNumberOfComponents(3)
vtkFloatArray->SetArray(&points[0], points.size() * 3, 1); // 1 means
vtkFloatArray->VTK does not free your array
vtkPoints->SetData(vtkFloatArray);
3. You can also pre-generate your vtkCellArray during initialization (and
extend it on-the-fly if necessary):
vector<vtkIdType> m_Indices;
m_Indices.reserve(2000000);
for (i = 0; i < 1000000; ++i)
{
m_Indices.push_back(1);
m_Indices.push_back(i);
}
Usage is like this:
vtkIdTypeArray->SetArray(&m_Indices[0], points.size() * 2, 1);
vtkCellArray->SetCells(points.size(), vtkIdTypeArray);
4. Create the VTK-pipeline
vtkPolyData->SetPoints(vtkPoints);
vtkPolyData->SetVerts(vtkCellArray);
vtkPolyDataMapper->SetInputConnection(0,
vtkPolyDataMapper->vtkPolyData->GetProducerPort());
vtkActor->SetMapper(vtkPolyDataMapper);
5. You can set point size through the properties of the vtkActor:
vtkActor->GetProperty()->SetPointSize(pointSize);
This is very fast and you should be able to render millions of points at a very
high frame rate. Maybe you can even get VTK to render your points without the
need of indices... Should be possible somehow I guess.
The points are rendered as squares which is kind of ugly. Solution: Add
mitkWorkbench to your AMD/NVidia application profiles and enable MSAA.
Best,
Stefan
-----Ursprüngliche Nachricht-----
Von: Clarkson, Matt [mailto:[email protected]]
Gesendet: Mittwoch, 8. Mai 2013 15:15
An: Engelhardt, Sandy
Cc: mitk-users
Betreff: Re: [mitk-users] Bug 14890: Very Slow Rendering of PointSets
Please can you remind me.. when was the ITKv4 commit?
Is it possible to upgrade MITK to include 8150+14932, without ITKv4?
However, I suspect that the problem also lies with the fact that for each point
a sphere is rendered, and the resolution is set to 20 in both latitude and
longitude.
Is there any way to litterally just render a single point for each point in the
point set?
We are using 2 million points, and 2 million spheres is a bit ...
optimistic. :-)
Thanks
M
On 8 May 2013, at 13:42, "Engelhardt, Sandy"
<[email protected]>
wrote:
Dear Matt,
the bug, which occurred in the latest release, has been fixed within the scope
of bug 8150. The merge of "GenerateDataForRenderer" and "GenerateData" methods
caused performance issues in a couple of other mappers too (see bug 14932).
Please let us know whether you still have problems with rendering large
pointsets.
Kind regards,
Sandy
-----Ursprüngliche Nachricht-----
Von: Clarkson, Matt [mailto:[email protected]]
Gesendet: Mittwoch, 8. Mai 2013 14:13
An: mitk-users
Betreff: [mitk-users] Bug 14890: Very Slow Rendering of PointSets
Hi there,
Does anyone have an update on this:
http://bugs.mitk.org/show_bug.cgi?id=14890
We are running into a problem rendering large pointsets, and so would be REALLY
interested to know if anyone had taken a look at this, or had suggestions.
Thanks
Matt
---------------------------------------------------------------------
-
-------- Learn Graph Databases - Download FREE O'Reilly Book "Graph
Databases" is the definitive new guide to graph databases and their
applications. This 200-page book is written by three acclaimed leaders in the
field. The early access version is available now.
Download your free book today! http://p.sf.net/sfu/neotech_d2d_may
_______________________________________________
mitk-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/mitk-users
----------------------------------------------------------------------
-------- Learn Graph Databases - Download FREE O'Reilly Book "Graph
Databases" is the definitive new guide to graph databases and their
applications. This 200-page book is written by three acclaimed leaders in the
field. The early access version is available now.
Download your free book today! http://p.sf.net/sfu/neotech_d2d_may
_______________________________________________
mitk-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/mitk-users
------------------------------------------------------------------------------
Learn Graph Databases - Download FREE O'Reilly Book
"Graph Databases" is the definitive new guide to graph databases and
their applications. This 200-page book is written by three acclaimed
leaders in the field. The early access version is available now.
Download your free book today! http://p.sf.net/sfu/neotech_d2d_may
_______________________________________________
mitk-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/mitk-users
