Custom Transforms

Custom Transforms—Test

Custom Transforms—Test

Anatomy of an Operator

def transform_scalars(dataset):

  from tomviz import utils
  import numpy as np

  # Get the current volume as a NumPy array.
  array = utils.get_array(dataset)

  # Operate on your data, here we square root it.
  result = np.sqrt(array)

  # Set the transformed data, displayed in Tomviz.
  utils.set_array(dataset, result)    

Classier Operators—Outline

import tomviz.operators

class SquareRootOperator(tomviz.operators.CancelableOperator):

  def transform_scalars(self, dataset):
    from tomviz import utils
    import numpy as np
    self.progress.maximum = 10
    scalars = utils.get_scalars(dataset)
    # Process dataset in chunks, user can cancel.
    result = np.float32(scalars)
    step = 0
    # Processing loop goes here...
    utils.set_scalars(dataset, result)

Classier Operators—Loop

def transform_scalars(self, dataset):
  # ...other stuff from previous slide
  scalars = utils.get_scalars(dataset)
  # Process dataset in chunks, user can cancel.
  result = np.float32(scalars)
  for chunk in np.array_split(result, 10):
    if self.canceled:
      return
    np.sqrt(chunk, chunk)
    step += 1
    self.progress.value = step

  utils.set_scalars(dataset, result)

Classier Operators—Complete

import tomviz.operators
CHUNKS = 10

class SquareRootOperator(tomviz.operators.CancelableOperator):

  def transform_scalars(self, dataset):
    from tomviz import utils
    import numpy as np
    self.progress.maximum = CHUNKS
    scalars = utils.get_scalars(dataset)
    # Process dataset in chunks, user can cancel.
    result = np.float32(scalars)
    step = 0
    for chunk in np.array_split(result, CHUNKS):
      if self.canceled:
        return
      np.sqrt(chunk, chunk)
      step += 1
      self.progress.value = step
    utils.set_scalars(dataset, result)

JSON for Increased Control

{
  "name": "Fancy Square Root",
  "label": "Classy Square Root",
  "description": "This is the fanciest square root operator, it does it all...",
  "parameters": [
    {
      "name": "number_of_chunks",
      "label": "Number of Chunks",
      "type": "int",
      "default": 10,
      "minimum": 1,
      "maximum": 1000
    }
  ]
}

Classier Operators—Complete

import tomviz.operators

class SquareRootOperator(tomviz.operators.CancelableOperator):

  def transform_scalars(self, dataset, number_of_chunks = 10):
    from tomviz import utils
    import numpy as np
    self.progress.maximum = number_of_chunks
    scalars = utils.get_scalars(dataset)
    # Process dataset in chunks, user can cancel.
    result = np.float32(scalars)
    step = 0
    for chunk in np.array_split(result, number_of_chunks):
      if self.canceled:
        return
      np.sqrt(chunk, chunk)
      step += 1
      self.progress.value = step
    utils.set_scalars(dataset, result)

Custom Transforms—JSON Menu

Custom Transforms—User Input

Custom Transforms—Result

Pipeline Settings

Pipeline Settings—Docker

Pulling Docker Image

Running Pipelines

• Create a virtual environment, install

  $ git clone --recursive git://github.com/openchemistry/tomviz
  $ cd tomviz/tomviz/python
  $ mkvirtualenv tomviz
  $ pip install -e .

• Use Tomviz to build a pipeline
  • Save the state file (JSON)

Running Pipelines

• Run the pipeline in the virtual environment

  $ tomviz-pipeline -s state.tvsm -d data.emd -o output.emd
  [2019-07-23 14:14:59,647] INFO: Executing 'Invert Data' operator
  [2019-07-23 14:14:59,963] INFO: Writing transformed data. 

• Input must be an EMD file right now
• Operators are executed in sequence
• Output is written to an EMD file