MclassDraw

Board	Supported
Host System	Yes
V4L2	Yes
Clarity UHD	Yes
Concord PoE	No
GenTL	Yes
GevIQ	Yes
GigE Vision	Yes
Indio	No
Iris GTX	Yes
Radient eV-CL	Yes
Rapixo CL	Yes
Rapixo CoF	Yes
Rapixo CXP	Yes
USB3 Vision	Yes

Draw specific features of a classifier context, a dataset context, or a prediction result buffer.

Syntax

void MclassDraw(
    AIL_ID    ContextGraId,            //in
    AIL_ID    ContextOrResultClassId,  //in
    AIL_ID    DstImageBufOrListGraId,  //out
    AIL_INT64 Operation,               //in
    AIL_INT64 Index,                   //in
    AIL_INT64 ControlFlag              //in
)

Description

This function draws specific features of a classifier context, a dataset context, or a prediction result buffer. Drawing operations are typically performed in a destination image buffer. You cannot draw from a tree ensemble result buffer, and you cannot perform any drawing operations with ONNX.

By default, this function performs the drawing operation for every class in the specified context or result. To draw for a specific class or specific instance of a class, specify its index (Index).

Parameters

ContextGraId (in, AIL_ID)

Specifies the identifier of the 2D graphics context to use when drawing. This parameter must be set to one of the following values:

For specifying the 2D graphics context

Value	Description
M_DEFAULT	Specifies that the default 2D graphics context of the current Aurora Imaging Library application is used.

Note: Note that there is a different default 2D graphics context for each thread. | | 2D graphics context identifier | Specifies a valid 2D graphics context identifier, which you have allocated using MgraAlloc. |

ContextOrResultClassId (in, AIL_ID)

Specifies the identifier of the classifier or dataset context, or prediction result buffer from which to draw.

For specifying the classifier or dataset context, or prediction result buffer

Value	Description
ContextClassifierPredefinedId	Specifies the identifier of the predefined CNN, object detection, segmentation, or anomaly detection classifier context. This context is allocated using MclassAlloc with M_CLASSIFIER_CNN_PREDEFINED,M_CLASSIFIER_DET_PREDEFINED, M_CLASSIFIER_SEG_PREDEFINED, or M_CLASSIFIER_ANO_PREDEFINED.
ContextClassifierTreeEnsembleId	Specifies the identifier of a tree ensemble classifier context. This context is allocated using MclassAlloc with M_CLASSIFIER_TREE_ENSEMBLE.
ContextDatasetId	Specifies the identifier of an images or features dataset context. These contexts must have been previously allocated on the required system using MclassAlloc with M_DATASET_IMAGES or M_DATASET_FEATURES.
ResultClassPredictId	Specifies the identifier of the CNN, object detection, segmentation, or anomaly detections, prediction result buffer. This buffer is allocated using MclassAllocResult with M_PREDICT_CNN_RESULT, M_PREDICT_DET_RESULT M_PREDICT_SEG_RESULT, or M_PREDICT_ANO_RESULT.

DstImageBufOrListGraId (out, AIL_ID)

Specifies the identifier of the destination image buffer or LUT buffer in which to draw. You should typically specify an image, which can be any valid Aurora Imaging Library image buffer allocated using MbufAlloc.... By drawing into a display's overlay buffer, you can also annotate an image non-destructively. To specify a LUT, you must use the M_DRAW_CLASS_COLOR_LUT drawing operation.

Operation (in, AIL_INT64)

Specifies the drawing operation to perform.

For a predefined (CNN, segmentation, or object detection) or tree ensemble classifier context, or a dataset context (images or features)

Value	Description
M_DRAW_CLASS_COLOR_LUT	Draws the color (LUT) associated to the class. This color is defined using MclassControl with M_CLASS_DRAW_COLOR.

When performing this operation, you must set the destination buffer (DstImageBufOrListGraId) to the identifier of an Aurora Imaging Library LUT buffer. To inquire the width of this LUT, use M_NUMBER_OF_CLASSES. | | M_DRAW_CLASS_ICON | Draws the icon image associated to the class. This image is defined using MclassControl with M_CLASS_ICON_ID.

When performing this operation, you must set the destination buffer (DstImageBufOrListGraId) to the identifier of an Aurora Imaging Library image buffer. To inquire the format of this buffer (DstImageBufOrListGraId), such as the X- and Y-size, use M_CLASS_ICON_ID (the dimension of the destination image in which you draw should be the same as the class description's icon image). |

For a CNN or segmentation prediction result buffer

Value	Description
M_DRAW_BEST_INDEX_CONTOUR_IMAGE	Draws the contours of the best classes. If you do not specify the M_PSEUDO_COLOR combination value, Aurora Imaging Library uses the result type M_BEST_INDEX_IMAGE_TYPE to determine the type of the image to use. If you specify M_PSEUDO_COLOR, the image type must be 3 bands, 8-bit unsigned.

Note, this drawing operation supports combination values from the draw table. The combination values that you can specify depend on the type of prediction result buffer that you are using:

• For CNN prediction results, you can specify M_NETWORK_OUTPUT_SIZE, M_PSEUDO_COLOR, and M_REPLICATE_BORDER.
• For segmentation prediction results, you can specify M_PSEUDO_COLOR. | | M_DRAW_BEST_INDEX_IMAGE | Draws an image using the index value of the best class at each point. If you do not specify the M_PSEUDO_COLOR combination value, Aurora Imaging Library uses the result type M_BEST_INDEX_IMAGE_TYPE to determine the type of the image to use. If you specify M_PSEUDO_COLOR, the image type must be 3 bands, 8-bit unsigned.

Note, this drawing operation supports combination values from the draw table. The combination values that you can specify depend on the type of prediction result buffer that you are using:

• For CNN prediction results, you can specify M_NETWORK_OUTPUT_SIZE, M_PSEUDO_COLOR, and M_REPLICATE_BORDER.
• For segmentation prediction results, you can specify M_PSEUDO_COLOR. | | M_DRAW_BEST_SCORE_IMAGE | Draws an image using the best score at each point. If you do not specify the M_PSEUDO_COLOR combination value, the image type must be float or unsigned. If you specify M_PSEUDO_COLOR, the image type must be 3 bands, 8-bit unsigned.

Note, this drawing operation supports combination values from the draw table. The combination values that you can specify depend on the type of prediction result buffer that you are using:

• For CNN prediction results, you can specify M_NETWORK_OUTPUT_SIZE, M_PSEUDO_COLOR, and M_REPLICATE_BORDER.
• For segmentation prediction results, you can specify M_PSEUDO_COLOR. | | M_DRAW_CLASS_SCORES | Draws (fills) an image using the scores of a specific class. This drawing operation is not possible when drawing general results; you must set the Index parameter to a specific class.

Note, this drawing operation supports combination values from the draw table. The combination values that you can specify depend on the type of prediction result buffer that you are using:

• For CNN prediction results, you can specify M_NETWORK_OUTPUT_SIZE and M_REPLICATE_BORDER. If you specify M_NETWORK_OUTPUT_SIZE, the drawing is done with the classification map size (the ControlFlag parameter must be set to its default). If you do not specify the M_NETWORK_OUTPUT_SIZE combination value, you can set the ControlFlag parameter to M_BILINEAR_INTERPOLATION or M_RECEPTIVE_FIELD_PATCH. In all cases, when using M_NETWORK_OUTPUT_SIZE, the image type must be float or 8-bit unsigned.
• For segmentation prediction results, you cannot specify any combination value.
• You cannot specify M_PSEUDO_COLOR for any predefined classifier. |

For an object detection prediction result buffer

Value	Description
M_DRAW_BOX	Draws the bounding boxes for all instances, a specific class, or a specific instance (only one box will be drawn in this case). The drawn color is specified in the class definitions. To change the line thickness of the box, you can use MgraControl with M_LINE_THICKNESS on the ContextGraId.
M_DRAW_BOX_CENTER	Draws the center positions of the bounding boxes for all instances, a specific class, or a specific instance (only one center position will be drawn in this case). The drawn color is specified in the class definitions. To change the line thickness of the center position, you can use MgraControl with M_LINE_THICKNESS with the graphics context specified by the ContextGraId parameter.
M_DRAW_BOX_NAME	Draws the class names for bounding boxes for all instances, a specific class, or a specific instance (only one name will be drawn in this case).

The class names are drawn using the graphics context specified by the ContextGraId parameter. You can modify the text settings of the drawn names using MgraControl.

If used with M_DRAW_BOX, the class name is drawn on top of the outline of the bounding box. If used with M_DRAW_BOX_CENTER, the class name is drawn near the center of the bounding box. Text alignment controls set with MgraControl are relative to these locations. | | M_DRAW_BOX_SCORE | Draws the class scores for bounding boxes for all instances, a specific class, or a specific instance (only one score will be drawn in this case).

The class scores are drawn using the graphics context specified by the ContextGraId parameter. You can modify the text settings of the drawn scores using MgraControl.

If used with M_DRAW_BOX, the class score is drawn on top of the outline of the bounding box. If used with M_DRAW_BOX_CENTER, the class score is drawn near the center of the bounding box. Text alignment controls set with MgraControl are relative to these locations. |

For an anomaly detection prediction result buffer

Value	Description
M_DRAW_ANOMALY_CONTOUR_MASK	Draws the contour of the anomalies using the color of the graphics context specified by the ContextGraId parameter. To change the color, you can use MgraControl with M_COLOR on the graphics context specified by the ContextGraId parameter.

Note, DstImageBufOrListGraId must be an image buffer of type M_UNSIGNED. Additionally, if the graphics context specified by the ContextGraId parameter is not grayscale, DstImageBufOrListGraId must be 3 bands and 8-bit. | | M_DRAW_ANOMALY_HEATMAP | Draws the anomaly scores projected as a heat map for each pixel location in the image.

Note, DstImageBufOrListGraId must be an 8-bit and 3-band image buffer of type M_UNSIGNED. | | M_DRAW_ANOMALY_MASK | Draws the anomalous pixels using the color of the graphics context specified by the ContextGraId parameter. To change the color, you can use MgraControl with M_COLOR on the graphics context specified by the ContextGraId parameter.

Note, DstImageBufOrListGraId must be an image buffer of type M_UNSIGNED. Additionally, if the graphics context specified by the ContextGraId parameter is not grayscale, DstImageBufOrListGraId must be 3 bands and 8-bit. | | M_DRAW_ANOMALY_SCORES | Draws the anomaly score for each pixel location in the image.

Note, DstImageBufOrListGraId must be a 1-band image buffer of type M_FLOAT. |

For modifying the drawing operation

Value	Description
M_NETWORK_OUTPUT_SIZE	Draws according to the exact classification map size (output of the network) without scaling or expanding. In this case, the expected size of the destination buffer (DstImageBufOrListGraId) is the same as the X- and Y-size of M_PREDICTION_SIZE_X and M_PREDICTION_SIZE_Y.

You cannot use M_NETWORK_OUTPUT_SIZE with M_REPLICATE_BORDER. | | M_PSEUDO_COLOR | Draws using the pseudo colors defined by calling MclassControl with M_CLASS_DRAW_COLOR. For all drawing operations, if you specify M_PSEUDO_COLOR, the destination buffer (DstImageBufOrListGraId) image type must be 3 bands, 8-bit unsigned.

Depending on the draw operation, you can use M_PSEUDO_COLOR with other combination values. | | M_REPLICATE_BORDER | Draws with borders expanded to the size of the destination buffer (DstImageBufOrListGraId).

You cannot use M_REPLICATE_BORDER with M_NETWORK_OUTPUT_SIZE. |

Index (in, AIL_INT64)

Specifies what to use to perform the drawing operation.

For specifying what to use to perform the drawing operation

Value	Description
M_DEFAULT	Same as M_CONTEXT, if the ContextOrResultClassId parameter is a predefined CNN, segmentation, object detection, or anomaly detection classifier context, a tree ensemble classifier context, an images dataset context, or a features dataset context; same as M_GENERAL, if the ContextOrResultClassId is a CNN, segmentation, object detection, or anomaly detection prediction result buffer.
M_CLASS_INDEX	Specifies that the drawing operation is for a specific class in the specified context or result buffer (ContextOrResultClassId).

You can draw for a specific class when specifying a CNN prediction result buffer (M_PREDICT_CNN_RESULT), an object detection prediction result buffer (M_PREDICT_DET_RESULT), a segmentation prediction result buffer (M_PREDICT_SEG_RESULT), a predefined CNN classifier context (M_CLASSIFIER_CNN_PREDEFINED), a predefined object detection classifier context (M_CLASSIFIER_DET_PREDEFINED), a predefined segmentation classifier context (M_CLASSIFIER_SEG_PREDEFINED), a tree ensemble classifier context (M_CLASSIFIER_TREE_ENSEMBLE), an images dataset context(M_DATASET_IMAGES), or a features dataset context (M_DATASET_FEATURES). | | M_INSTANCE_INDEX | Specifies that the drawing operation is for a specific instance in the specified object detection prediction result buffer (ContextOrResultClassId). | | M_CONTEXT | Specifies that the drawing operation is for every class in the classifier context or the dataset context (ContextOrResultClassId). | | M_GENERAL | Specifies that the drawing operation is for every class in the CNN, segmentation, object detection, or anomaly detection prediction result buffer (ContextOrResultClassId). |

ControlFlag (in, AIL_INT64)

Specifies how to control the drawing operation.

For specifying how to control the class score drawing

Value	Description
M_DEFAULT	Implements the default behavior. This is the only possible setting for all drawing operations except M_DRAW_CLASS_SCORES.
M_BILINEAR_INTERPOLATION	Scales the classification map with a bilinear calculation, when using M_DRAW_CLASS_SCORES. The scale is established in receptive field calculations. The scaled image is placed at the offset of the receptive field.
M_RECEPTIVE_FIELD_PATCH	Fills the corresponding receptive fields with calculated scores, when using M_DRAW_CLASS_SCORES. Scores are drawn in order, from low to high. The best score is therefore drawn over lower scores.