Skip to content

Latest commit

 

History

History
79 lines (46 loc) · 4.26 KB

paraview.md

File metadata and controls

79 lines (46 loc) · 4.26 KB
Raw

FTK ParaView Guide

Limitations

FTK's features are partially implemented in ParaView plug-ins. We are addressing the following limitations for future releases:

  • We do not currently support distributed and parallel processing with ParaView plug-ins. One can compile FTK with MPI and use mpiexec to execute ftk in order to track features with multiple processes
  • We have limited support for CUDA-accelerated filters. One can compile FTK with CUDA and use --accelerator cuda to accelerate computation with a GPU
  • Performance of our filters (especially 4D filters) may not be optimized because of MPI/CUDA support is missing
  • Filters are only supported for vtkImageData types for now; use C++ APIs for unstructured mesh data
  • Post-processing of outputs, e.g. smoothing, simplification, and filtering of critical point trajectories not yet supported in ParaView plug-ins

Installation

Currently, the only way to install FTK's ParaView plug-ins is to build from source. See install.md for more details.

FTK synthetic data sources

See synthetic.md for a comprehensive list of synthetic data sources that are supported

FTK Filters

CriticalPointTracker2D

This filter tracks critical points in 2D scalar and vector fields.

Inputs

The input must be in 2D vtkImageData type. We currently do not support vtkUnstructuredGrid as the input.

The filter automatically determines whether the input is a scalar or vector field based on the number of components of the input variable:

  • The input is treated as a scalar field if the input variable contains one single component. For example, the scalar variable provided by MovingExtremum2DSource is a scalar field because the variable contains one single component
  • The input is treated as a vector field if the input variable contains multiple components. For example, the vector variable provided by the DoubleGyre2DSource is a vector field because the variable contains more than one components
  • If a vector field is provided by two different variables, say u and v, one has to combine the two variables into a multi-component variable using the Calculator filter with iHat*u + jHat*v; otherwise u or v will be treated as a scalar field.

Options

  • InputVariable specifies the input variable
  • ZTimeScale (by default zero) controls the ratio of z-coordinates over time in the outputs, which can be used to generate spacetime visualizations in 3D (x and y being space and z being time). This option is by default zero in order to behave consistently as CriticalPointTracker3D, which is not possible to visualize spacetime visualizations in 3D

Outputs

The outputs are trajectories in vtkPolyData format with multiple variables:

  • id are the integer identifiers of trajectories
  • time are the time coordinates of each point
  • type is the type of critical points. See ftk/numeric/critical_point_type.hh for definitions

CriticalPointTracker3D

This filter tracks critical points in 3D scalar and vector fields. We omit descriptions because this filter behaves very similar to CriticalPointTracker2D.

LevelsetTracker2D

This filter tracks level sets (contours) in 2D scalar fields.

Inputs

The input must be in 2D vtkImageData type; the input variable must contain one single component. We currently do not support vtkUnstructuredGrid as the input.

Options

  • Threshold (by default zero) is the the isovalue for levelest tracking
  • OutputType controls whether the outputs are "sliced" (contours in individual timesteps) or "traced" (spacetime isosurfaces); see the above image
  • ZTimeScale (by default zero) controls the ratio of z-coordinates over time in the outputs; see the ZTimeScale in CriticalPointTracker2D for more details

Outputs

The "sliced" outputs are contours in the vtkPolyData format for each tilmestep; the "traced" outputs are surfaces in the vtkPolyData format. The outputs have multiple variables:

  • id are the integer identifiers of trajectories
  • time are the time coordinates of each point

LevelSetTracker3D

This filter tracked levelests (contours) in 3D scalar fields. We omit descriptions because this filter behaves very similar to LevelSetTracker3D.