add_samples_image_class_svm🔗
Short description🔗
add_samples_image_class_svm — Add training samples from an image to the training data of a support
vector machine.
Signature🔗
add_samples_image_class_svm( image Image, region ClassRegions, class_svm SVMHandle )
Description🔗
add_samples_image_class_svm adds training samples from the
image Image to the support vector machine (SVM) given by
SVMHandle. add_samples_image_class_svm is used to
store the training samples before training a classifier for the
pixel classification of multichannel images with
classify_image_class_svm.
add_samples_image_class_svm works analogously to
add_sample_class_svm.
The image Image must have a number of channels equal to
NumFeatures, as specified with create_class_svm. The
training regions for the NumClasses pixel classes are passed in
ClassRegions. Hence, ClassRegions must be a tuple
containing NumClasses regions. The order of the regions in
ClassRegions determines the class of the pixels. If there
are no samples for a particular class in Image, an empty
region must be passed at the position of the class in
ClassRegions. With this mechanism it is possible to use
multiple images to add training samples for all relevant classes to
the SVM by calling add_samples_image_class_svm multiple
times with the different images and suitably chosen regions.
The regions in ClassRegions should contain representative
training samples for the respective classes. Hence, they need not
cover the entire image. The regions in ClassRegions should
not overlap each other, because this would lead to the fact that in
the training data the samples from the overlapping areas would be
assigned to multiple classes, which may lead to slower convergence
of the training and a lower classification performance.
A further application of this operator is the automatic novelty detection,
where, e.g., anomalies in color or texture can be detected. For this
mode a training set that defines a sample region (e.g., skin
regions for skin detection or samples of the correct texture) is
passed to the SVMHandle, which is created in the Mode
'novelty-detection'. After training, regions that differ from the
trained sample regions are detected (e.g., the rejection class for skin
or errors in texture).
Execution information🔗
Execution information
-
Multithreading type: reentrant (runs in parallel with non-exclusive operators).
-
Multithreading scope: global (may be called from any thread).
-
Processed without parallelization.
This operator modifies the state of the following input parameter:
During execution of this operator, access to the value of this parameter must be synchronized if it is used across multiple threads.
Parameters🔗
Image (input_object) (multichannel-)image → object (byte / cyclic / direction / int1 / int2 / uint2 / int4 / real)
Training image.
ClassRegions (input_object) region-array → object
Regions of the classes to be trained.
SVMHandle (input_control, state is modified) class_svm → (handle)
SVM handle.
Result🔗
If the parameters are valid add_samples_image_class_svm
returns the value 2 (H_MSG_TRUE). If necessary, an exception is
raised.
Combinations with other operators🔗
Combinations
Possible predecessors
Possible successors
train_class_svm, write_samples_class_svm
Alternatives
See also
classify_image_class_svm, add_sample_class_svm, clear_samples_class_svm, get_sample_num_class_svm, get_sample_class_svm, add_samples_image_class_mlp
Module🔗
Foundation