[vtkusers] Help with VtkDatasetsurface filter

Alex Malyushytskyy alexmalvtk at gmail.com
Tue Apr 2 17:49:07 EDT 2013


LOD (level of details)  technique consists of run time assessment of
performance and building a special model which will be used in certain
circumstances and on the one hand would look as close as possible to
the original, but on another hand allows achieve desired or better
performance.
That what LOD actor does, crude models will be silently created and
displayed during rotation or when actor is moved.

Now assume there are 2 model - lets say 1st is your unstructured grid,
another is result of dataSetSurfaceFilter.
Since both are using LOD actors, the following might happen:

1st model is found to be too big to be displayed and as a result crude
model with N1 number of nodes is created and displayed
2nd model is found to be too big to be displayed and as a result crude
model with N2 number of nodes is created and displayed
or it is small enough , so it is displayed as is (but lets say still
have N2 nodes)

Now N1 might be mostly the same as N2.
And you see no difference in performance during transition or rotation.
Basically this means that LOD actor does his job well enough.
I bet if you were using regular actors you would see a huge
performance increase.

If you need further performance you may just specify different frame rate.
>From http://www.vtk.org/doc/release/4.0/html/classvtkLODActor.html:

"To control the frame rate, you typically set the
vtkRenderWindowInteractor DesiredUpdateRate and StillUpdateRate. This
then will cause vtkLODActor to adjust its LOD to fulfill the requested
update rate.
For greater control on levels of detail, see also vtkLODProp3D. That
class allows arbitrary definition of each LOD.
"

Hope this helps,

Alex




On Tue, Apr 2, 2013 at 2:18 PM, Bhanu Hariharan <bhanu at petrotel.com> wrote:
> Alex,
>
> To be specific in one of my cases,  No. of surfaces rendered were reduced
> from 6163542 to 373626 assuming grid has hexahedrons each with 6 surfaces. I
> got the number 373626 by doing
>
> dataSetSurfaceFilter->GetOutput()->GetNumberOfCells()
>
> Thats a significant reduction of number of surfaces. These are rendered
> using VTKLODActor.
>
> Can you please explain what you mean by "actually shown primitives during
> rotation". What are the shown primitives?
>
> I am just adding the output of the dataSetSurfaceFilter to a
> vtkDataSetMapper and then to a vtkLODActor.
>
> All I know if no. of surfaces are reduced. Do I need to do/check something
> else?
>
> If you need something specific to answer more specifically, I can provide
> the details. It would be a great help.
>
> Thanks,
> Bhanu
>
>
> ---------- Forwarded message ----------
> From: Alex Malyushytskyy <alexmalvtk at gmail.com>
> Date: Tue, Apr 2, 2013 at 4:03 PM
> Subject: Re: [vtkusers] Help with VtkDatasetsurface filter
> To: "vtkusers at vtk.org" <vtkusers at vtk.org>
>
>
> You may get only general advice asking general questions.
> Compare  the number of nodes/cells you are you are getting from
> vtkDataSetSurfaceFilter with
> such in the unstructured grid,
> Even though it may be significantly decreased in some cases, for
> example if LOD actors are used,  they may decrease actually shown
> primitives during rotation approximately to the same number and you
> will not see any difference in the performance.
>
> Regards,
>     Alex
>
>
> On Tue, Apr 2, 2013 at 8:27 AM, Bhanu Hariharan <bhanu at petrotel.com> wrote:
>> Hi All,
>> I was wondering if anyone would be kind enough to answer the following
>> query
>> for me.
>>
>> I had posted this a while back and still havent found the reason why
>> vtkDataSetSurfaceFilter hasn't made rendering any faster. The interaction
>> and rendering time to rotate zoom etc seem to be just the same as though I
>> rendered all cells in the unstructured grid.
>> The reason I chose to introduce rendering just surface cells in the first
>> place was to make rendering faster. But I am surprised it doesnt make any
>> difference at all :(.
>> Any words or ideas?
>>
>> Thanks so much
>> Bhanu
>>
>>
>>
>> On Mon, Mar 11, 2013 at 4:30 PM, Bhanu Hariharan <bhanu at petrotel.com>
>> wrote:
>>>
>>> One related question though. I started using Surface Filter really only
>>> for performance reasons.
>>>
>>> I was thinking if I rendered just surface cells, the rendering and
>>> rotation etc would be must faster.
>>> However I do not find it any different than when I render all cells. What
>>> am I missing?
>>> Any reasons anyone can think of, please let me know.
>>>
>>>
>>>
>>> Thanks,
>>> Bhanu
>>>
>>>
>>>
>>> On Mon, Mar 11, 2013 at 4:12 PM, Bhanu Hariharan <bhanu at petrotel.com>
>>> wrote:
>>>>
>>>> David,
>>>> As you suggested, I got it to work using SetGlobalIds and then
>>>> retrieving
>>>> that using GetGlobalIds. Many thanks for your patience and help.
>>>>
>>>>
>>>> -Bhanu
>>>>
>>>>
>>>> On Wed, Mar 6, 2013 at 11:41 AM, Bhanu Hariharan <bhanu at petrotel.com>
>>>> wrote:
>>>>>
>>>>> David,
>>>>>
>>>>> I am using vtk 5.2 - is it possible that this function was not defined
>>>>> in that version.
>>>>>
>>>>> I get the error that SetOriginalCellIdsName is not a member of
>>>>> vtkDataSetSurfaceFilter   :(
>>>>>
>>>>> Bhanu
>>>>>
>>>>>
>>>>>
>>>>> On Wed, Mar 6, 2013 at 10:36 AM, David E DeMarle
>>>>> <dave.demarle at kitware.com> wrote:
>>>>>>
>>>>>> Give DSF a name for the array.
>>>>>> SetOriginalCellIdsName("idsintoDSForsomething")
>>>>>> Then retrieve it from the output like
>>>>>> DSF->GetOutput()->GetCellData()->GetArray("idsintoDSForsomething")
>>>>>>
>>>>>>
>>>>>> David E DeMarle
>>>>>> Kitware, Inc.
>>>>>> R&D Engineer
>>>>>> 21 Corporate Drive
>>>>>> Clifton Park, NY 12065-8662
>>>>>> Phone: 518-881-4909
>>>>>>
>>>>>>
>>>>>> On Wed, Mar 6, 2013 at 11:29 AM, Bhanu Hariharan <bhanu at petrotel.com>
>>>>>> wrote:
>>>>>>>
>>>>>>>
>>>>>>> David,
>>>>>>>
>>>>>>> I had turned on PassThroughCell. But I do not know how to access the
>>>>>>> new Ids array after the fact.
>>>>>>> VTK Doc says something about a 'OrginalCellIds' but I am not sure how
>>>>>>> to access this.
>>>>>>> So I extracted the output of DSF to a vtkPolyData but still cant
>>>>>>> figure out how to get to that new Ids array.
>>>>>>> Any help with that is really appreciated.
>>>>>>>
>>>>>>> I will try your other suggestion in the meantime.
>>>>>>>
>>>>>>> Thanks,
>>>>>>> Bhanu
>>>>>>>
>>>>>>>
>>>>>>> On Wed, Mar 6, 2013 at 9:25 AM, David E DeMarle
>>>>>>> <dave.demarle at kitware.com> wrote:
>>>>>>>>
>>>>>>>> Sorry for the confusion, it has been a long time since I added that
>>>>>>>> code the DataSetSurfaceFilter. Turning PassThroughCell/PointIds in
>>>>>>>> fact
>>>>>>>> generates a new Ids array, which you can retrieve after the fact to
>>>>>>>> do the
>>>>>>>> lookups into the data that is input to the DSF and extract
>>>>>>>> particular
>>>>>>>> cells/points by their offset (n'th cell in the cell array, n'th
>>>>>>>> point in the
>>>>>>>> point array)
>>>>>>>>
>>>>>>>> However I still recommend making your own ids array and using that
>>>>>>>> instead of PassThrough. vtkIdFilter will generate them for you. The
>>>>>>>> array
>>>>>>>> marked as the active global ids array is treated specially by most
>>>>>>>> filters
>>>>>>>> in that they try to preserve it across operations (including
>>>>>>>> threshold and
>>>>>>>> DSF). We added this to VTK a few years ago specifically for this
>>>>>>>> case of
>>>>>>>> finding out where particular result values come from (and in
>>>>>>>> particular for
>>>>>>>> surface selection). So the Threshold and surface filter will push
>>>>>>>> globalIDS
>>>>>>>> forward just fine. Once you get a result, grab the active globaIDs
>>>>>>>> array,
>>>>>>>> look for the values in it, and look them up in the original data.
>>>>>>>> The
>>>>>>>> extract selections filter will do the job of looking them up in the
>>>>>>>> original
>>>>>>>> data and pulling them out as a new dataset.
>>>>>>>>
>>>>>>>>
>>>>>>>> David E DeMarle
>>>>>>>> Kitware, Inc.
>>>>>>>> R&D Engineer
>>>>>>>> 21 Corporate Drive
>>>>>>>> Clifton Park, NY 12065-8662
>>>>>>>> Phone: 518-881-4909
>>>>>>>>
>>>>>>>>
>>>>>>>> On Wed, Mar 6, 2013 at 10:02 AM, Bhanu Hariharan
>>>>>>>> <bhanu at petrotel.com>
>>>>>>>> wrote:
>>>>>>>>>
>>>>>>>>> David,
>>>>>>>>> Thanks for replying. However I have a doubt.
>>>>>>>>> Are you suggesting the VTKUnstructuredgrid should have an globalids
>>>>>>>>> array? Is that what the member is called - globalIds? I actually
>>>>>>>>> pass
>>>>>>>>> VTKUnstructuredgrid  to a VTKThreshold filter and then pass that
>>>>>>>>> filter to
>>>>>>>>> VTKdatasetsurfacefilter.
>>>>>>>>>
>>>>>>>>> Perhaps you can post a couple lines of code. That would be very
>>>>>>>>> helpful.
>>>>>>>>> Thanks,
>>>>>>>>> Bhanu
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> On Tue, Mar 5, 2013 at 6:55 PM, David E DeMarle
>>>>>>>>> <dave.demarle at kitware.com> wrote:
>>>>>>>>>>
>>>>>>>>>> The input data should have a globalids array containing a unique
>>>>>>>>>> value for each cell. If it does, extract will preserve it. After
>>>>>>>>>> the
>>>>>>>>>> extraction, use that array in the polydata to do the look up to
>>>>>>>>>> the original
>>>>>>>>>> cell in the unstructured grid.
>>>>>>>>>>
>>>>>>>>>> On Mar 5, 2013 6:00 PM, "Bhanu Hariharan" <bhanu at petrotel.com>
>>>>>>>>>> wrote:
>>>>>>>>>>>
>>>>>>>>>>> Hi All,
>>>>>>>>>>>
>>>>>>>>>>> I need some help doing the following. Please could one of you
>>>>>>>>>>> guide me on how.
>>>>>>>>>>>
>>>>>>>>>>> I use a vtkDataSetSurfaceFilter on VTKUnstructuredGrid to render
>>>>>>>>>>> skin of that grid. That works fine and I can see the number of
>>>>>>>>>>> surfaces
>>>>>>>>>>> shown by grabbing the
>>>>>>>>>>> dataSetSurfaceFilter->GetOutput()->GetNumberOfCells().
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> I need a way to map the surface back to the original cell id.
>>>>>>>>>>> That
>>>>>>>>>>> is if I have 2 cells and I render just skin, number of surfaces
>>>>>>>>>>> rendered are
>>>>>>>>>>> 10. I need to be able to pick on any of these 10 surfaces and map
>>>>>>>>>>> them back
>>>>>>>>>>> to either of the the 2 grid cells.
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> http://www.vtk.org/doc says :
>>>>>>>>>>>
>>>>>>>>>>> If PassThroughCellIdsOn is 1, then the output polygonal dataset
>>>>>>>>>>> will have a celldata array that holds the cell index of the
>>>>>>>>>>> original 3D cell
>>>>>>>>>>> that produced each output cell. This is useful for cell picking.
>>>>>>>>>>> I turned the PassThroughCellIds On and I have UseStrips turned
>>>>>>>>>>> off.
>>>>>>>>>>>
>>>>>>>>>>> Then I did
>>>>>>>>>>> vtkPolyData *surfacePolyData = dataSetSurfaceFilter->GetOutput();
>>>>>>>>>>>
>>>>>>>>>>> But I do not see how I can access that celldata array that holds
>>>>>>>>>>> the original 3D cells for the surfaces.
>>>>>>>>>>>
>>>>>>>>>>> Please can anyone advise on how.
>>>>>>>>>>>
>>>>>>>>>>> Thanks in advance.
>>>>>>>>>>> Bhanu
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> _______________________________________________
>>>>>>>>>>> Powered by www.kitware.com
>>>>>>>>>>>
>>>>>>>>>>> Visit other Kitware open-source projects at
>>>>>>>>>>> http://www.kitware.com/opensource/opensource.html
>>>>>>>>>>>
>>>>>>>>>>> Please keep messages on-topic and check the VTK FAQ at:
>>>>>>>>>>> http://www.vtk.org/Wiki/VTK_FAQ
>>>>>>>>>>>
>>>>>>>>>>> Follow this link to subscribe/unsubscribe:
>>>>>>>>>>> http://www.vtk.org/mailman/listinfo/vtkusers
>>>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>
>>>>>
>>>>
>>>
>>
>>
>> _______________________________________________
>> Powered by www.kitware.com
>>
>> Visit other Kitware open-source projects at
>> http://www.kitware.com/opensource/opensource.html
>>
>> Please keep messages on-topic and check the VTK FAQ at:
>> http://www.vtk.org/Wiki/VTK_FAQ
>>
>> Follow this link to subscribe/unsubscribe:
>> http://www.vtk.org/mailman/listinfo/vtkusers
>>
> _______________________________________________
> Powered by www.kitware.com
>
> Visit other Kitware open-source projects at
> http://www.kitware.com/opensource/opensource.html
>
> Please keep messages on-topic and check the VTK FAQ at:
> http://www.vtk.org/Wiki/VTK_FAQ
>
> Follow this link to subscribe/unsubscribe:
> http://www.vtk.org/mailman/listinfo/vtkusers
>



More information about the vtkusers mailing list