Using simplification on ContourForests results in crash

[Twinklebear]

I'm trying to use the ContourForests code and would like to simplify the tree after computing it via the simplification threshold stuff. However, when trying to simplify the tree getSeedsPos eventually returns size_t(-1) as the first tuple element, which the merge tree code then tries to retrieve the super arc on this overflowed id. At this point the code crashes with [Merge Tree] get superArc on bad id :18446744073709551615 / 129 (I'm running on the fuel from https://github.com/pavolzetor/open_scivis_datasets).

Breaking in the debugger it looks like this occurs here, where mergingArcId is set to the overflowed negative id.

I'm using ttk through the vtk wrappers, here's the code to reproduce:

// includes....

int main(int argc, char **argv) {
    // Read the volume data using vtk
    vtkSmartPointer<vtkXMLImageDataReader> reader
        = vtkSmartPointer<vtkXMLImageDataReader>::New();
    reader->SetFileName(args[1].c_str());
    reader->Update();
    vtkImageData *vol = reader->GetOutput();
    assert(vol);
    std::cout << "loaded img '" << args[1] << "'\n";

    vtkSmartPointer<vtkContourForests> contourForest
        = vtkSmartPointer<vtkContourForests>::New();
    contourForest->SetInputData(reader->GetOutput());
    contourForest->SetTreeType(ttk::TreeType::Split);
    contourForest->SetArcResolution(20);
    contourForest->SetSkeletonSmoothing(50);
    contourForest->SetUseAllCores(true);
    contourForest->SetThreadNumber(12);
    contourForest->SetSimplificationType(ttk::SimplifMethod::Persist);
    contourForest->SetSimplificationThreshold(0.5);
    contourForest->SetLessPartition(true);
    contourForest->SetDebugLevel(2);
    contourForest->Update();
    return 0;
}

The defines withParallelSimplify and withOpenMP also seem to need to be set. Maybe for another bug report, but without these no simplification is run. From a glance it looks like #defines in ContourForests::parallelBuild here and here have no fallback if withParallelSimplify is not set.

[julien-tierny]

Hi Will, thanks for your email.

I'm trying to use the ContourForests code and would like to simplify the tree after computing it via the simplification threshold stuff.

This feature is no longer supported (we're going to remove it from the API soon). Topological simplification in TTK should be used as described in the companion technical report: https://topology-tool-kit.github.io/stuff/ttk.pdf In short, you should select the extrema you want to maintain (with any heuristic of your choice), call the topological simplification and then compute the abstraction you're interested in (for instance: critical points, persistence diagrams, merge trees, contour trees or Morse-Smale complexes) Please check the following videos to see how to do this in ParaView: https://youtu.be/YVk9vRKIEX8 https://youtu.be/xjKh6YTq5RA To replicate this in Python or VTK/C++, you just need to replicate the ParaView pipeline shown in the above videos. Such a transcription is showcased in Figure 1 of the TTK technical report (the corresponding code is attached to this email, just change the Morse-Smale complex for the contour forest plugin). Thanks for letting us know if you run into any other issue! By the way, you may consider registering to the user mailing list (by sending an empty email to ttk-users+subscribe@googlegroups.com) to participate to similar usage discussions in the future. Cheers, -- Dr Julien Tierny CNRS Researcher Sorbonne Universites UPMC - LIP6 http://lip6.fr/Julien.Tierny

…

On Thursday 20 April 2017 8:02:34 PM Will Usher wrote: I'm trying to use the ContourForests code and would like to simplify the tree after computing it via the simplification threshold stuff. However, when trying to simplify the tree `getSeedsPos` eventually returns `size_t(-1)` as the first tuple element, which the merge tree code then tries to retrieve the super arc on this overflowed id. At this point the code crashes with `[Merge Tree] get superArc on bad id :18446744073709551615 / 129` (I'm running on the fuel from https://github.com/pavolzetor/open_scivis_datasets). Breaking in the debugger it looks like this occurs [here](https://github.com/topology-tool-kit/ttk/blob/master/core/baseCode/c ontourForestsTree/MergeTreeTemplate.h#L277), where mergingArcId is set to the overflowed negative id. I'm using ttk through the vtk wrappers, here's the code to reproduce: ```cpp // includes.... int main(int argc, char **argv) { // Read the volume data using vtk vtkSmartPointer<vtkXMLImageDataReader> reader = vtkSmartPointer<vtkXMLImageDataReader>::New(); reader->SetFileName(args[1].c_str()); reader->Update(); vtkImageData *vol = reader->GetOutput(); assert(vol); std::cout << "loaded img '" << args[1] << "'\n"; vtkSmartPointer<vtkContourForests> contourForest = vtkSmartPointer<vtkContourForests>::New(); contourForest->SetInputData(reader->GetOutput()); contourForest->SetTreeType(ttk::TreeType::Split); contourForest->SetArcResolution(20); contourForest->SetSkeletonSmoothing(50); contourForest->SetUseAllCores(true); contourForest->SetThreadNumber(12); contourForest->SetSimplificationType(ttk::SimplifMethod::Persist); contourForest->SetSimplificationThreshold(0.5); contourForest->SetLessPartition(true); contourForest->SetDebugLevel(2); contourForest->Update(); return 0; } ``` The defines `withParallelSimplify` and `withOpenMP` also seem to need to be set. Maybe for another bug report, but without these no simplification is run. From a glance it looks like `#defines` in `ContourForests::parallelBuild` [here](https://github.com/topology-tool-kit/ttk/blob/master/core/baseCode/c ontourForests/ContourForestsTemplate.h#L341) and [here](https://github.com/topology-tool-kit/ttk/blob/master/core/baseCode/c ontourForests/ContourForestsTemplate.h#L366) have no fallback if `withParallelSimplify` is not set. -- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: #9

[Twinklebear]

Thanks Julien, the code you suggested works great 👍

[Twinklebear]

Hey @julien-tierny I ran into one small issue with the contour forests using this, in the example code you pasted it looks like vtk requires that you explicitly set the offset scalar field name to use, otherwise the simplification doesn't seem right? It seemed to do some sort of simplification but it wouldn't match what I'd get in paraview with the same settings, setting the scalar field name explicitly has got it to match paraview, which looks to set the name automatically

[julien-tierny]

hi Will, thanks for your message.

Hey @julien-tierny I ran into one small issue with the contour forests using this, in the example code you pasted it looks like vtk requires that you explicitly set the offset scalar field name to use,

That's my bad. The default values for these fields were correctly set only for the Morse-Smale complex (which I precisely tuned for the purpose of the code demo shown in Figure 1 of TTK's tech report). I've just pushed some edits regarding the default values of these fields (in the VTK wrappers and ParaView plugins) for all scalar data modules. The contour forests can now be used exactly like the Morse-Smale complex of the demo code I sent you (in principle). Thanks for letting us know if you run into any other issue! Cheers, -- Dr Julien Tierny CNRS Researcher Sorbonne Universites UPMC - LIP6 http://lip6.fr/Julien.Tierny

…

On Monday 24 April 2017 4:21:13 PM Will Usher wrote: Hey @julien-tierny I ran into one small issue with the contour forests using this, in the example code you pasted it looks like vtk requires that you explicitly set the offset scalar field name to use, otherwise the simplification doesn't seem right? It seemed to do some sort of simplification but it wouldn't match what I'd get in paraview with the same settings, setting the scalar field name explicitly has got it to match paraview, which looks to set the name automatically -- You are receiving this because you were mentioned. Reply to this email directly or view it on GitHub: #9 (comment)