CClassificationTask#

class ikomia.dataprocess.pydataprocess.CClassificationTask#

Base class for classification task in Computer Vision. It defines a task with the following properties:

Inputs:

image (CImageIO)
graphics (CGraphicsInput)

Outputs:

image IO (CImageIO): by default source image is forwarded.
object detection IO (CObjectDetectionIO): filled if input graphics items are passed. In this case, classification is computed for each individual object.
graphics output (CGraphicsOutput): text item with top-1 class if classification is computed on whole image.
data output (CDataStringIO): sorted list of class scores if classification is computed on whole image.

Derived from C2dImageTask.

Import

from ikomia.dataprocess import CClassificationTask

Methods

`__init__`(arg1)	__init__( (object)self) -> None :
`add_object`(self, graphics_item, class_index, ...)	Add classification result for individual object.
`get_input_objects`(self)	Get input graphics items on which classification can be computed individually.
`get_names`(self)	Get class names.
`get_objects_results`(self)	Get classification results when applied on individual objects (input graphics items).
`get_object_sub_image`(self, item)	Get ROI image for the given graphics item.
`get_image_with_graphics`(self)	Get visualization image where all extracted information are embedded (graphics items).
`get_whole_image_results`(self)	Get classification results when applied on whole image (no input graphics items given).
`is_whole_image_classification`(self)	Check whether input graphics items are given for individual classification.
`read_class_names`(self, path)	Populate class names from the given text file (one line per class).
`set_colors`(self, colors)	Set colors associated with class names.
`set_names`(self, names)	Set class names.
`set_whole_image_results`(self, names, confidences)	Set whole image classification results.
`emit_add_sub_progress_steps`(self, count)	See `emit_add_sub_progress_steps()`.
`emit_graphics_context_changed`(self)	See `emit_graphics_context_changed()`.
`emit_output_changed`(self)	See `emit_output_changed()`.
`emit_step_progress`(self)	See `emit_step_progress()`.

Overridden methods

end_task_run(self)

See end_task_run().

Inherited methods

`add_input`(self, input)	Add new input to the task.
`add_output`(self, output)	Add new output to the task.
`apply_graphics_mask`(self, origin, processed, ...)	Apply the mask generated from graphics to the result image so that only masked areas seems to be processed.
`apply_graphics_mask_to_binary`(self, origin, ...)	Apply the mask generated from graphics to the binary source image.
`begin_task_run`(self)	See `begin_task_run()`.
`create_input_graphics_mask`(self, index, ...)	Generate a binary mask image from the task graphics input at the specified index.
`create_graphics_mask`(self, width, height, ...)	Generate a binary mask image from the given graphics input object.
`execute_actions`(self, action)	Method called when a specific action is requested from the associated widget (see `emit_send_process_action()`).
`forward_input_image`(self, input_index, ...)	Forward input image at position input_index to output at position output_index.
`get_elapsed_time`(self)	Get the time of the last execution in milliseconds.
`get_graphics_mask`(self, index)	Get the binary mask generated from graphics input at position index.
`get_input`(self, index)	Get input at position index.
`get_input_count`(self)	Get the number of inputs.
`get_input_data_type`(self, index)	Get input data type at position index.
`get_inputs`(self)	Get the whole list of inputs.
`get_output`(self, index)	Get output at position index.
`get_output_count`(self)	Get the number of outputs.
`get_output_data_type`(self, index)	Get output data type at position index.
`get_outputs`(self)	Get the whole list of outputs.
`get_parameters`(self)	Get values of task parameters.
`get_progress_steps`(self)	See `get_progress_steps()`.
`global_input_changed`(self, is_new_sequence)	See `global_input_changed()`.
`graphics_changed`(self)	See `graphics_changed()`.
`is_graphics_changed_listening`(self)	Check whether the task is listening to graphics changed event.
`is_mask_available`(self, index)	Check whether a binary mask from graphics input is available at position index.
`parameters_modified`(self)	Notify that the task parameters have changed.
`remove_input`(self, index)	Remove input at the given position.
`run`(self)	See `run()`.
`set_action_flag`(self, action, is_enable)	Enable or disable the given action.
`set_active`(self, is_active)	See `set_active()`.
`set_input`(self, input, index)	Set input at position index with the given one.
`set_input_data_type`(self, data_type, index)	Set the data type for the input at position index.
`set_inputs`(self, inputs)	Set the whole list of inputs with the given one.
`set_output`(self, output, index)	Set output at position index with the given one.
`set_output_color_map`(self, index, ...)	Bind a display color map to an image output.
`set_output_data_type`(self, data_type, index)	Set the data type for the output at position index.
`set_outputs`(self, outputs)	Set the whole list of outputs with the given one.
`set_parameters`(self, values)	Set values of task parameters.
`stop`(self)	See `stop()`.
`update_static_outputs`(self)	Determine output data type automatically from input data types.

Attributes

`name`	Task name (must be unique)
`type`	Main purpose or data type on which the task is dedicated to.

Details

__init__((object)arg1) → None#

__init__( (object)self) -> None :

Default constructor

__init__( (object)self, (str)name) -> None :

Construct CClassificationTask object with the given name.

Args:

name (str): task name, must be unique

add_object((CClassificationTask)self, (CGraphicsItem)graphics_item, (int)class_index, (float)confidence) → None :#

Add classification result for individual object. See get_input_objects() and get_object_sub_image() for more information.

Parameters:

graphics_item (CGraphicsItem based object) –
class_index (int) – index is used to retrieve class name
confidence (float) – confidence score of top-1 class

get_input_objects((CClassificationTask)self) → object :#

Get input graphics items on which classification can be computed individually. One can iterate over this list to compute classification for each object. Use get_object_sub_image() to retieve object ROI image and add_object() to store classification result.

Returns:: graphics items
Return type:: CGraphicsItem based objects

get_names((CClassificationTask)self) → object :#

Get class names. Call read_class_names() to populate names from text file.

Returns:: class names
Return type:: str list

get_objects_results((CClassificationTask)self) → CObjectDetectionIO :#

Get classification results when applied on individual objects (input graphics items). Results are given as a CObjectDetectionIO instance.

Returns:: classification results
Return type:: CObjectDetectionIO

get_object_sub_image((CClassificationTask)self, (CGraphicsItem)item) → object :#

Get ROI image for the given graphics item. We use the bounding rect property of CGraphicsItem to compute ROI. Input graphics items can be retrieved with get_input_objects(). Classification can then be computed on the ROI to get individual object class.Args:

graphics item (CGraphicsItem based object)

Returns:: ROI image
Return type:: 2D Numpy array

get_image_with_graphics((CClassificationTask)self) → object :#

Get visualization image where all extracted information are embedded (graphics items).

Returns:: visualization image
Return type:: 2D Numpy array

get_whole_image_results((CClassificationTask)self) → object :#

Get classification results when applied on whole image (no input graphics items given). It gives a sorted list of tuple storing class name and confidence.

Returns:: classification results
Return type:: list of tuples (name, confidence)

is_whole_image_classification((CClassificationTask)self) → bool :#

Check whether input graphics items are given for individual classification.

Returns:: True if no input graphics items are given (whole image classification), False otherwise
Return type:: bool

read_class_names((CClassificationTask)self, (str)path) → None :#

Populate class names from the given text file (one line per class).

Parameters:: path (str) – path to class names definition file

set_colors((CClassificationTask)self, (object)colors) → None :#

Set colors associated with class names. The given list must have the same size as names list. If not provided, random colors are generated while populating the name list (read_class_names()).

Parameters:: list (colors (list of) – r, g, b integer values in range [0, 255])

set_names((CClassificationTask)self, (object)names) → None :#

Set class names. The function generate associated random colors if none is defined.

Parameters:: names (list of str) –

set_whole_image_results((CClassificationTask)self, (object)names, (object)confidences) → None :#

Set whole image classification results.

Parameters:

names (str list) – sorted list with respect to confidence score
confidences (str list) – sorted list (descending)

emit_add_sub_progress_steps((CClassificationTask)self, (int)count) → None :#: See emit_add_sub_progress_steps().

emit_graphics_context_changed((CClassificationTask)self) → None :#: See emit_graphics_context_changed().

emit_output_changed((CClassificationTask)self) → None :#: See emit_output_changed().

emit_step_progress((CClassificationTask)self) → None :#: See emit_step_progress().

end_task_run((CClassificationTask)self) → None :#

See end_task_run().

end_task_run( (CClassificationTask)self) -> None