MclassGetResult

Retrieves an array containing the development dataset accuracy for each epoch.

Note that the required array size, which you can retrieve with M_NB_ELEMENTS, is equivalent to the M_MAX_EPOCH (MclassControl()) of the training context used.

Retrieves an array containing the development dataset error rate for each epoch.

Note that the required array size, which you can retrieve with M_NB_ELEMENTS, is equivalent to the M_MAX_EPOCH (MclassControl()) of the training context used.

Retrieves an array containing the MIL_UUID of each image in the development dataset that could not be loaded, at least once, during training.

Retrieves the index of the epoch where classifier parameters were updated the last time.

Retrieves the number of mini-batches per epoch.

Retrieves an array containing the training dataset accuracy for each epoch.

Note that the required array size, which you can retrieve with M_NB_ELEMENTS, is equivalent to the M_MAX_EPOCH (MclassControl()) of the training context used.

Retrieves an array containing the training dataset error rate for each epoch.

Note that the required array size, which you can retrieve with M_NB_ELEMENTS, is equivalent to the M_MAX_EPOCH (MclassControl()) of the training context used.

Retrieves an array containing the MIL_UUID of each image in the training dataset that could not be loaded, at least once, during training.

Retrieves an array containing the loss error values for each mini-batch during training. You can use the loss to evaluate the lack of confidence (doubt) associated with the classification during training.

Note that the required array size, which you can retrieve with M_NB_ELEMENTS, is equivalent to M_MAX_EPOCH (MclassControl()) * M_MINI_BATCH_PER_EPOCH (MclassGetResult()) of the training context used.

Retrieves an array containing the accuracies on the development dataset after adding each tree to the classifier.

Retrieves an array containing the error rates after on the development dataset after adding each tree to the classifier.

Retrieves an array containing the importance of each feature when training the classifier.

This result can be disabled if M_FEATURE_IMPORTANCE_MODE is set to M_DISABLE.

Retrieves the number of out-of-bag dataset entries.

This is only available if M_COMPUTE_OUT_OF_BAG_RESULTS is set to M_ENABLE and M_BOOTSTRAP is not set to M_DISABLE.

Retrieves the number of features.

Retrieves the number of trained trees. This result is equivalent to the number of trees in the classifier.

Retrieves the out-of-bag error accuracy.

This is only available if M_COMPUTE_OUT_OF_BAG_RESULTS is set to M_ENABLE and M_BOOTSTRAP is not set to M_DISABLE.

Retrieves the array containing the accuracies on the out-of-bag set after adding each tree to the classifier.

This is only available if M_COMPUTE_OUT_OF_BAG_RESULTS is set to M_ENABLE and M_BOOTSTRAP is not set to M_DISABLE.

Retrieves the confusion matrix obtained using the out-of-bag set.

This is a square matrix. MIL returns the matrix values, row-by-row, as an array. The array size is equivalent to the square of the total amount of classes, which is the same M_OUT_OF_BAG_CONFUSION_MATRIX_SIZE_X x M_OUT_OF_BAG_CONFUSION_MATRIX_SIZE_Y, or M_NB_ELEMENTS.

This is only available if M_COMPUTE_OUT_OF_BAG_RESULTS is set to M_ENABLE and M_BOOTSTRAP is not set to M_DISABLE.

Retrieves the X-dimension of the confusion matrix. This value is calculated using the out-of-bag set.

This is only available if M_COMPUTE_OUT_OF_BAG_RESULTS is set to M_ENABLE and M_BOOTSTRAP is not set to M_DISABLE.

Retrieves the Y-dimension of the confusion matrix. This value is calculated using the out-of-bag set.

This is only available if M_COMPUTE_OUT_OF_BAG_RESULTS is set to M_ENABLE and M_BOOTSTRAP is not set to M_DISABLE.

Retrieves the out-of-bag error rate.

This is only available if M_COMPUTE_OUT_OF_BAG_RESULTS is set to M_ENABLE and M_BOOTSTRAP is not set to M_DISABLE.

Retrieves the array containing the accuracies on the out-of-bag set after adding each tree to the classifier.

This is only available if M_COMPUTE_OUT_OF_BAG_RESULTS is set to M_ENABLE and M_BOOTSTRAP is not set to M_DISABLE.

Retrieves the proximity matrix.

This is a square matrix. MIL returns the matrix values, row-by-row, as an array. The array size is equivalent M_PROXIMITY_MATRIX_SIZE_X x M_PROXIMITY_MATRIX_SIZE_Y or M_NB_ELEMENTS.

In the proximity matrix, the entry in cell (j, k) is some measure of similarity (or distance) between the entries to which row j and column k correspond.

The proximity matrix result can be enabled or disabled using M_COMPUTE_PROXIMITY_MATRIX in the train context.

Retrieves the X-dimension of the proximity matrix.

Retrieves the Y-dimension of the proximity matrix.

Retrieves the seed used for the training dataset.

Retrieves the array containing the accuracies on the training dataset after adding each tree to the classifier.

Retrieves the array containing the error rates on the training dataset after adding each tree to the classifier.

Retrieves the array containing the depth achieved by each tree in the classifier.

Retrieves the array containing the number of leaf nodes for each tree in the classifier.

Retrieves the accuracy on the development dataset using the trained context.

Retrieves the confusion matrix obtained using the development dataset.

This is a square matrix. MIL returns the matrix values, row-by-row, as an array. The array size is equivalent to the square of the total amount of classes, which is the same M_DEV_DATASET_CONFUSION_MATRIX_SIZE_X x M_DEV_DATASET_CONFUSION_MATRIX_SIZE_Y, or M_NB_ELEMENTS.

Retrieves the X-dimension of the confusion matrix. This value is calculated using the development dataset.

Retrieves the Y-dimension of the confusion matrix. This value is calculated using the development dataset.

Retrieves the error rate on the development dataset. This can be expressed as 100.0 - M_DEV_DATASET_ACCURACY.

Retrieves an array containing the MIL_UUID of each image in the development dataset used during the training phase.

Retrieves the accuracy on the training dataset using the trained context.

Retrieves the confusion matrix obtained using the training dataset.

This is a square matrix. MIL returns the matrix values, row-by-row, as an array. The array size is equivalent to the square of the total amount of classes, which is the same M_TRAIN_DATASET_CONFUSION_MATRIX_SIZE_X x M_TRAIN_DATASET_CONFUSION_MATRIX_SIZE_Y, or M_NB_ELEMENTS.

Retrieves the X-dimension of the confusion matrix. This value is calculated using the training dataset.

Retrieves the Y-dimension of the confusion matrix. This value is calculated using the training dataset.

Retrieves the error rate on the training dataset. This can be expressed as 100.0 - M_TRAIN_DATASET_ACCURACY.

Retrieves an array containing the MIL_UUID of each image in the training dataset used during the training phase.

Retrieves the image type to provide when calling MclassDraw() with M_DRAW_BEST_INDEX_IMAGE.

Retrieves the classification map offset along the X-axis.

Apply the scale and offset to M_CLASSIFICATION_MAP_SIZE_X to compute the center of the receptive field of each M_CLASS_SCORES into the target image such as: x * M_CLASSIFICATION_MAP_SCALE_X + M_CLASSIFICATION_MAP_OFFSET_X.

Retrieves the classification map offset along the Y-axis.

Apply the scale and offset to M_CLASSIFICATION_MAP_SIZE_Y to compute the center of the receptive field of each M_CLASS_SCORES into the target image such as: y * M_CLASSIFICATION_MAP_SCALE_Y + M_CLASSIFICATION_MAP_OFFSET_Y.

Retrieves the classification map scale along the X-axis.

Apply the scale and offset to M_CLASSIFICATION_MAP_SIZE_X to compute the center of the receptive field of each CLASS_SCORES into the target image such as: x * M_CLASSIFICATION_MAP_SCALE_X + M_CLASSIFICATION_MAP_OFFSET_X.

Retrieves the classification map scale along the Y-axis.

Apply the scale and offset to M_CLASSIFICATION_MAP_SIZE_Y to compute the center of the receptive field of each CLASS_SCORES into the target image such as: y * M_CLASSIFICATION_MAP_SCALE_Y + M_CLASSIFICATION_MAP_OFFSET_Y.

Retrieves the number of predictions the deep network has made according to the target image size (see M_TARGET_IMAGE_SIZE_X).

Typically, for a target image the same size as the network's source layer the map size will be 1x1, meaning a single prediction for the whole target image. For a larger target image, the map size can be larger, meaning the CNN has convolved the larger target image and made many predictions doing so.

Retrieves the number of predictions the deep network has made according to the target image size (see M_TARGET_IMAGE_SIZE_Y).

Typically, for a target image the same size as the network's source layer the map size will be 1x1, meaning a single prediction for the whole target image. For a larger target image, the map size can be larger, meaning the CNN has convolved the larger target image and made many predictions doing so.

Retrieves the size of the receptive field along the X-axis.

Retrieves the size of the receptive field along the Y-axis.

Retrieves the highest class index.

Note, for coarse segmentation (pixel level classification) of an image with a CNN, this retrieves an array (map) of values that can be arranged in a 2D image, where the size is equal to M_CLASSIFICATION_MAP_SIZE_X multiplied by M_CLASSIFICATION_MAP_SIZE_Y. Each value is the index of the class having the best score. The index of the class begins at 0.

Retrieves the highest class score.

Retrieves the number of classes.

For a CNN, this refers to the number of outputs in the CNN's last layer. For tree ensemble, this refers to the number of tree ensemble in the classifier.

This can also be used in a tree ensemble prediction with LabelOrIndex set to M_GENERAL.

Retrieves an array of all the class scores.

The number of returned values is M_NUMBER_OF_CLASS_SCORES.

For general results obtained from a CNN prediction result buffer, the returned values are organized planar-wise in a 3d volume of size M_CLASSIFICATION_MAP_SIZE_X * M_CLASSIFICATION_MAP_SIZE_Y * M_NUMBER_OF_CLASSES. Note, each band contains the score of one given class. The index in this returned array of the score of class C at pixel X, Y is: index = (C * MapSizeX * MapSizeY) + (Y * MapSizeX) + X.

For general results obtained from a tree ensemble prediction result buffer, the number of class scores is equivalent to M_NUMBER_OF_CLASS_SCORES.

For class-specific results (LabelOrIndex set to M_CLASS_INDEX()) obtained from a CNN prediction result buffer, the returned values are organized in a vector that corresponds to a 2D volume of size M_CLASSIFICATION_MAP_SIZE_X * M_CLASSIFICATION_MAP_SIZE_Y. The index of the score at location X, Y is: index = (Y * M_CLASSIFICATION_MAP_SIZE_X ) + X.

Retrieves the total number of class scores.

For general results obtained from a CNN predict result buffer, the total corresponds to M_CLASSIFICATION_MAP_SIZE_X * M_CLASSIFICATION_MAP_SIZE_Y * M_NUMBER_OF_CLASSES scores. The first batch of M_CLASSIFICATION_MAP_SIZE_X * M_CLASSIFICATION_MAP_SIZE_Y scores are for the first class and the subsequent batches are for the remaining classes. In other words, the data is organized planar-wise.

For general results obtained from a tree ensemble predict result buffer, the total corresponds to M_NUMBER_OF_CLASSES.

For class-specific results (LabelOrIndex set to M_CLASS_INDEX()) obtained from a CNN predict result buffer, the total corresponds to M_CLASSIFICATION_MAP_SIZE_X * M_CLASSIFICATION_MAP_SIZE_Y.