Package org.anchoranalysis.plugin.image.task.bean.segment

Class SegmentInstanceWithModel<T extends InferenceModel>

Object

AnchorBean<Task<StackSequenceInput,SharedStateSegmentInstance<T>>>

Task<StackSequenceInput,SharedStateSegmentInstance<T>>

SegmentInstanceWithModel<T>

Type Parameters:

T - model-type in pool

public class SegmentInstanceWithModel<T extends InferenceModel>
extends Task<StackSequenceInput,SharedStateSegmentInstance<T>>

Using a model-pool, performs instance segmentation on an image producing zero, one or more objects per image.

Various visualizations and export types are supported.

• Segmentation results in vairous forms (HDF5) for input into other scripts. (in HDF5 form and as a mask).
• Visualizations of the instances found (outlines, thumbnails etc.)
• A table of basic features (CSV) for each instance.

Specifically, the following outputs are produced:

Output Name	Default?	Description
"input"	no	The input image for segmentation.
"objects"	yes	Segmented object-masks encoded into HDF5.
"mask"	yes	A binary-mask image that binary ors each voxel across the segmented object-masks (scaled to match the input image for model inference).
"outline"	yes	A RGB image showing the outline of segmented-objects on top of the input image (scaled to match the input image for model inference).
"mask""InputScale"	no	Like mask but on the full-scale input image.
"outline""InputScale"	no	Like outline but on on the full-scale input image.
"summary"	yes	A CSV file showing basic feature of all segmented-objects across all input images.
"thumbnails"<S>	yes	A directory of thumbnails showing the outline of all segmented objects on top of an extracted portion of the respective input-image.
outputs inherited from Task

Author:

Owen Feehan

Constructor Summary

Constructors

Constructor	Description
SegmentInstanceWithModel()

Method Summary

Modifier and Type	Method	Description
void	afterAllJobsAreExecuted(SharedStateSegmentInstance<T> sharedState, InputOutputContext context)	Called once after all calls to Task.executeJob(org.anchoranalysis.experiment.task.ParametersUnbound<T, S>).
SharedStateSegmentInstance<T>	beforeAnyJobIsExecuted(Outputter outputter, ConcurrencyPlan plan, List<StackSequenceInput> inputs, ParametersExperiment parameters)	Called once before all calls to Task.executeJob(org.anchoranalysis.experiment.task.ParametersUnbound<T, S>).
OutputEnabledMutable	defaultOutputs()	If specified, default rules for determine which outputs are enabled or not.
void	doJobOnInput(InputBound<StackSequenceInput,SharedStateSegmentInstance<T>> input)	Performs the task on a particular input.
List<NamedBean<FeatureListProvider<FeatureInputSingleObject>>>	getFeatures()	Features to calculate for objects in the features output.
Interpolator	getInterpolator()	The interpolator to use for scaling images.
RGBColorBean	getOutlineColor()	The color of the outline.
int	getOutlineWidth()	The width of the outline.
SegmentStackIntoObjectsPooled<T>	getSegment()	The segmentation algorithm.
ExportFeaturesStyle	getStyle()	Visual style for how feature export occurs.
boolean	hasVeryQuickPerInputExecution()	Is the execution-time of the task per-input expected to be very quick to execute?
InputTypesExpected	inputTypesExpected()	Highest class(es) that will function as a valid input.
boolean	isIgnoreNoObjects()	When true, then the outputs (outline, mask, image etc.) are not written for images that produce no objects.
boolean	isVaryColors()	When true, the colors change for different objects in the image (using a default color set).
void	setFeatures(List<NamedBean<FeatureListProvider<FeatureInputSingleObject>>> features)	Features to calculate for objects in the features output.
void	setIgnoreNoObjects(boolean ignoreNoObjects)	When true, then the outputs (outline, mask, image etc.) are not written for images that produce no objects.
void	setInterpolator(Interpolator interpolator)	The interpolator to use for scaling images.
void	setOutlineColor(RGBColorBean outlineColor)	The color of the outline.
void	setOutlineWidth(int outlineWidth)	The width of the outline.
void	setSegment(SegmentStackIntoObjectsPooled<T> segment)	The segmentation algorithm.
void	setStyle(ExportFeaturesStyle style)	Visual style for how feature export occurs.
void	setVaryColors(boolean varyColors)	When true, the colors change for different objects in the image (using a default color set).

Methods inherited from class org.anchoranalysis.experiment.bean.task.Task

executeJob, isInputCompatibleWith

Methods inherited from class org.anchoranalysis.bean.AnchorBean

checkMisconfigured, describeBean, describeChildren, duplicateBean, fields, findFieldsOfClass, getBeanName, getLocalPath, localise, toString

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait

Constructor Details

SegmentInstanceWithModel

public SegmentInstanceWithModel()

Method Details

inputTypesExpected

public InputTypesExpected inputTypesExpected()

Description copied from class: Task

Highest class(es) that will function as a valid input.

This is usually the class of T (or sometimes the absolute base class InputFromManager).

Specified by:

inputTypesExpected in class Task<StackSequenceInput,SharedStateSegmentInstance<T extends InferenceModel>>

Returns:

the input-types that a task expects.

beforeAnyJobIsExecuted

public SharedStateSegmentInstance<T> beforeAnyJobIsExecuted(Outputter outputter,
 ConcurrencyPlan plan,
 List<StackSequenceInput> inputs,
 ParametersExperiment parameters)
                                                     throws ExperimentExecutionException

Description copied from class: Task

Called once before all calls to Task.executeJob(org.anchoranalysis.experiment.task.ParametersUnbound<T, S>).

Specified by:

beforeAnyJobIsExecuted in class Task<StackSequenceInput,SharedStateSegmentInstance<T extends InferenceModel>>

Parameters:

outputter - the output-manager for the experiment (not for an individual job)

plan - available numbers of processors that can call Task.executeJob(org.anchoranalysis.experiment.task.ParametersUnbound<T, S>)

inputs - a list of inputs, each will result in at least one call to Task.executeJob(ParametersUnbound).

parameters - the experiment-parameters.

Returns:

the shared-state that is passed to each call to Task.executeJob(org.anchoranalysis.experiment.task.ParametersUnbound<T, S>) and to Task.afterAllJobsAreExecuted(S, org.anchoranalysis.io.output.outputter.InputOutputContext).

Throws:

ExperimentExecutionException - if a fatal error occurs executing this step of the experiment.

doJobOnInput

public void doJobOnInput(InputBound<StackSequenceInput,SharedStateSegmentInstance<T>> input)
                  throws JobExecutionException

Description copied from class: Task

Performs the task on a particular input.

Specified by:

doJobOnInput in class Task<StackSequenceInput,SharedStateSegmentInstance<T extends InferenceModel>>

Parameters:

input - the input for the job.

Throws:

JobExecutionException - if an error occurs executing a particular job, that is not otherwise suppressed.

afterAllJobsAreExecuted

public void afterAllJobsAreExecuted(SharedStateSegmentInstance<T> sharedState,
 InputOutputContext context)
                             throws ExperimentExecutionException

Description copied from class: Task

Called once after all calls to Task.executeJob(org.anchoranalysis.experiment.task.ParametersUnbound<T, S>).

Specified by:

afterAllJobsAreExecuted in class Task<StackSequenceInput,SharedStateSegmentInstance<T extends InferenceModel>>

Parameters:

sharedState - the shared-state

context - IO-context for experiment (not for an individual job)

Throws:

ExperimentExecutionException - if a fatal error occurs executing this step of the experiment.

hasVeryQuickPerInputExecution

public boolean hasVeryQuickPerInputExecution()

Description copied from class: Task

Is the execution-time of the task per-input expected to be very quick to execute?

Specified by:

hasVeryQuickPerInputExecution in class Task<StackSequenceInput,SharedStateSegmentInstance<T extends InferenceModel>>

Returns:

true iff the execution is expected to be very quick, false otherwise.

defaultOutputs

public OutputEnabledMutable defaultOutputs()

Description copied from class: Task

If specified, default rules for determine which outputs are enabled or not.

Overrides:

defaultOutputs in class Task<StackSequenceInput,SharedStateSegmentInstance<T extends InferenceModel>>

Returns:

the default rules if they exist.

getSegment

public SegmentStackIntoObjectsPooled<T> getSegment()

The segmentation algorithm.

setSegment

public void setSegment(SegmentStackIntoObjectsPooled<T> segment)

The segmentation algorithm.

getOutlineColor

public RGBColorBean getOutlineColor()

The color of the outline.

setOutlineColor

public void setOutlineColor(RGBColorBean outlineColor)

The color of the outline.

isVaryColors

public boolean isVaryColors()

When true, the colors change for different objects in the image (using a default color set).

This takes precedence over outlineColor.

setVaryColors

public void setVaryColors(boolean varyColors)

When true, the colors change for different objects in the image (using a default color set).

This takes precedence over outlineColor.

getOutlineWidth

public int getOutlineWidth()

The width of the outline.

setOutlineWidth

public void setOutlineWidth(int outlineWidth)

The width of the outline.

getFeatures

public List<NamedBean<FeatureListProvider<FeatureInputSingleObject>>> getFeatures()

Features to calculate for objects in the features output.

If unspecified, default features of bounding-box coordinates and number of voxels are selected.

setFeatures

public void setFeatures(List<NamedBean<FeatureListProvider<FeatureInputSingleObject>>> features)

Features to calculate for objects in the features output.

If unspecified, default features of bounding-box coordinates and number of voxels are selected.

isIgnoreNoObjects

public boolean isIgnoreNoObjects()

When true, then the outputs (outline, mask, image etc.) are not written for images that produce no objects.

setIgnoreNoObjects

public void setIgnoreNoObjects(boolean ignoreNoObjects)

When true, then the outputs (outline, mask, image etc.) are not written for images that produce no objects.

getStyle

public ExportFeaturesStyle getStyle()

Visual style for how feature export occurs.

setStyle

public void setStyle(ExportFeaturesStyle style)

Visual style for how feature export occurs.

getInterpolator

public Interpolator getInterpolator()

The interpolator to use for scaling images.

setInterpolator

public void setInterpolator(Interpolator interpolator)

The interpolator to use for scaling images.