set_dl_classifier_paramT_set_dl_classifier_paramSetDlClassifierParamSetDlClassifierParam (Operator)

Name

set_dl_classifier_paramT_set_dl_classifier_paramSetDlClassifierParamSetDlClassifierParam — Set the parameters of a deep-learning-based classifier.

Signature

set_dl_classifier_param( : : DLClassifierHandle, GenParamName, GenParamValue : )

Herror T_set_dl_classifier_param(const Htuple DLClassifierHandle, const Htuple GenParamName, const Htuple GenParamValue)

void SetDlClassifierParam(const HTuple& DLClassifierHandle, const HTuple& GenParamName, const HTuple& GenParamValue)

void HDlClassifier::SetDlClassifierParam(const HTuple& GenParamName, const HTuple& GenParamValue) const

void HDlClassifier::SetDlClassifierParam(const HString& GenParamName, const HString& GenParamValue) const

void HDlClassifier::SetDlClassifierParam(const char* GenParamName, const char* GenParamValue) const

void HDlClassifier::SetDlClassifierParam(const wchar_t* GenParamName, const wchar_t* GenParamValue) const   (Windows only)

static void HOperatorSet.SetDlClassifierParam(HTuple DLClassifierHandle, HTuple genParamName, HTuple genParamValue)

void HDlClassifier.SetDlClassifierParam(HTuple genParamName, HTuple genParamValue)

void HDlClassifier.SetDlClassifierParam(string genParamName, string genParamValue)

Description

set_dl_classifier_paramset_dl_classifier_paramSetDlClassifierParamSetDlClassifierParamSetDlClassifierParam sets the parameters and hyperparameters GenParamNameGenParamNameGenParamNameGenParamNamegenParamName of the neural network DLClassifierHandleDLClassifierHandleDLClassifierHandleDLClassifierHandleDLClassifierHandle to the values GenParamValueGenParamValueGenParamValueGenParamValuegenParamValue.

The pretrained classifiers are trained for their default image dimensions, see read_dl_classifierread_dl_classifierReadDlClassifierReadDlClassifierReadDlClassifier.

The network architectures allow different image dimensions. But for networks with at least one fully connected layer such a change makes a retraining necessary. Networks without fully connected layers are directly applicable to different image sizes. However, images with a size differing from the size with which the classifier has been trained, are likely to show a reduced classification accuracy.

GenParamNameGenParamNameGenParamNameGenParamNamegenParamName can attain the following values:

'batch_size'"batch_size""batch_size""batch_size""batch_size":
Number of images (and corresponding labels) in a batch and thus the number of images that are processed simultaneously in a single iteration of the training. Please refer to train_dl_classifier_batchtrain_dl_classifier_batchTrainDlClassifierBatchTrainDlClassifierBatchTrainDlClassifierBatch for further details. This parameter is stored in the pretrained classifier. Per default, the 'batch_size'"batch_size""batch_size""batch_size""batch_size" is set such that a training of the pretrained classifier with up to 100 classes can be easily performed on a device with 8 gigabyte of memory. For the pretrained classifiers, the default values are hence given as follows:
pretrained classifier default value of 'batch_size'"batch_size""batch_size""batch_size""batch_size" and 'batch_size_device'"batch_size_device""batch_size_device""batch_size_device""batch_size_device"
'pretrained_dl_classifier_compact.hdl'"pretrained_dl_classifier_compact.hdl""pretrained_dl_classifier_compact.hdl""pretrained_dl_classifier_compact.hdl""pretrained_dl_classifier_compact.hdl" '160'"160""160""160""160"
'pretrained_dl_classifier_enhanced.hdl'"pretrained_dl_classifier_enhanced.hdl""pretrained_dl_classifier_enhanced.hdl""pretrained_dl_classifier_enhanced.hdl""pretrained_dl_classifier_enhanced.hdl" '96'"96""96""96""96"
'pretrained_dl_classifier_resnet50.hdl'"pretrained_dl_classifier_resnet50.hdl""pretrained_dl_classifier_resnet50.hdl""pretrained_dl_classifier_resnet50.hdl""pretrained_dl_classifier_resnet50.hdl" '23'"23""23""23""23"
If you set 'batch_size'"batch_size""batch_size""batch_size""batch_size" to a value smaller than 'batch_size_device'"batch_size_device""batch_size_device""batch_size_device""batch_size_device", the latter one will automatically be adapted to 'batch_size'"batch_size""batch_size""batch_size""batch_size".
'batch_size_device'"batch_size_device""batch_size_device""batch_size_device""batch_size_device":

Number of images in the batch that is transferred to device memory. If this value is smaller than 'batch_size'"batch_size""batch_size""batch_size""batch_size" the current training batch is subdivided into smaller subbatches of size 'batch_size_device'"batch_size_device""batch_size_device""batch_size_device""batch_size_device". The gradient used for the update step is calculated by iterating over the subbatches. This enables arbitrarily large values of 'batch_size'"batch_size""batch_size""batch_size""batch_size", then although determined iteratively, the gradient used for the update is calculated taking into account all images of the batch. Note, 'batch_size_device'"batch_size_device""batch_size_device""batch_size_device""batch_size_device" can take values up to the maximum of 'batch_size'"batch_size""batch_size""batch_size""batch_size". Hence, it is not allowed to set 'batch_size_device'"batch_size_device""batch_size_device""batch_size_device""batch_size_device" to a value larger than 'batch_size'"batch_size""batch_size""batch_size""batch_size". If 'batch_size'"batch_size""batch_size""batch_size""batch_size" is set to a value smaller than 'batch_size_device'"batch_size_device""batch_size_device""batch_size_device""batch_size_device", the latter one will automatically be adapted to 'batch_size'"batch_size""batch_size""batch_size""batch_size". For the pretrained classifiers, the default values of 'batch_size_device'"batch_size_device""batch_size_device""batch_size_device""batch_size_device" are identical to the default values of 'batch_size'"batch_size""batch_size""batch_size""batch_size".

'classes'"classes""classes""classes""classes":

Tuple of labels corresponding to the classes of objects which are to be recognized. The order of the class names remains unchanged after the setting.

'gpu'"gpu""gpu""gpu""gpu":

Identifier of the GPU where the training and inference operators (train_dl_classifier_batchtrain_dl_classifier_batchTrainDlClassifierBatchTrainDlClassifierBatchTrainDlClassifierBatch and apply_dl_classifierapply_dl_classifierApplyDlClassifierApplyDlClassifierApplyDlClassifier) are executed. Per default, the first available GPU is used. get_systemget_systemGetSystemGetSystemGetSystem with 'cuda_devices'"cuda_devices""cuda_devices""cuda_devices""cuda_devices" can be used to retrieve a list of available GPUs. Pass the index in this list to 'gpu'"gpu""gpu""gpu""gpu".

'image_width'"image_width""image_width""image_width""image_width":

Width of the images the network will process. The default value is given by the network, see read_dl_classifierread_dl_classifierReadDlClassifierReadDlClassifierReadDlClassifier.

'image_height'"image_height""image_height""image_height""image_height":

Height of the images the network will process. The default value is given by the network, see read_dl_classifierread_dl_classifierReadDlClassifierReadDlClassifierReadDlClassifier.

'image_num_channels'"image_num_channels""image_num_channels""image_num_channels""image_num_channels":

Number of channels of the images the network will process. Possible are one channel (gray value image), or three channels (three-channel image). The default value is given by the network, see read_dl_classifierread_dl_classifierReadDlClassifierReadDlClassifierReadDlClassifier. Changing to a single channel image modifies the network configuration. This process removes the color information contained in certain layers and is not invertible.

'image_dimensions'"image_dimensions""image_dimensions""image_dimensions""image_dimensions":

Tuple containing the image dimensions 'image_width'"image_width""image_width""image_width""image_width", 'image_height'"image_height""image_height""image_height""image_height", and number of channels 'image_num_channels'"image_num_channels""image_num_channels""image_num_channels""image_num_channels". The default values are given by the network, see read_dl_classifierread_dl_classifierReadDlClassifierReadDlClassifierReadDlClassifier. Concerning the number of channels, the values one (gray value image), or three (three-channel image) are possible. Changing to a single channel image modifies the network configuration. This process removes the color information contained in certain layers and is not invertible.

'learning_rate'"learning_rate""learning_rate""learning_rate""learning_rate":

Initial value of the factor determining the gradient influence during training. Please refer to train_dl_classifier_batchtrain_dl_classifier_batchTrainDlClassifierBatchTrainDlClassifierBatchTrainDlClassifierBatch for further details. Per default, the 'learning_rate'"learning_rate""learning_rate""learning_rate""learning_rate" is set to '0.001'"0.001""0.001""0.001""0.001".

'momentum'"momentum""momentum""momentum""momentum":

When updating the arguments of the loss function, the hyperparameter 'momentum'"momentum""momentum""momentum""momentum" specifies to which extent previous updating vectors will be added to the current updating vector. Please refer to train_dl_classifier_batchtrain_dl_classifier_batchTrainDlClassifierBatchTrainDlClassifierBatchTrainDlClassifierBatch for further details. Per default, the 'momentum'"momentum""momentum""momentum""momentum" is set to '0.9'"0.9""0.9""0.9""0.9".

'runtime'"runtime""runtime""runtime""runtime":

Defines the device on which the operators will be executed. Per default, the 'runtime'"runtime""runtime""runtime""runtime" is set to 'gpu'"gpu""gpu""gpu""gpu".

'cpu'"cpu""cpu""cpu""cpu":

The operator apply_dl_classifierapply_dl_classifierApplyDlClassifierApplyDlClassifierApplyDlClassifier will be executed on the CPU, whereas the operator train_dl_classifier_batchtrain_dl_classifier_batchTrainDlClassifierBatchTrainDlClassifierBatchTrainDlClassifierBatch is not executable.

In case the GPU has been used before, CPU memory is initialized, and if necessary values stored on the GPU memory are moved to the CPU memory.

The 'cpu'"cpu""cpu""cpu""cpu" runtime uses OpenMP for the parallelization of apply_dl_classifierapply_dl_classifierApplyDlClassifierApplyDlClassifierApplyDlClassifier. Per default, all threads available to the OpenMP runtime are used. Use the thread specific set_systemset_systemSetSystemSetSystemSetSystem parameter 'tsp_thread_num'"tsp_thread_num""tsp_thread_num""tsp_thread_num""tsp_thread_num" to specify the number of threads to use.

'gpu'"gpu""gpu""gpu""gpu":

The GPU memory is initialized and the corresponding handle created. The operators apply_dl_classifierapply_dl_classifierApplyDlClassifierApplyDlClassifierApplyDlClassifier and train_dl_classifier_batchtrain_dl_classifier_batchTrainDlClassifierBatchTrainDlClassifierBatchTrainDlClassifierBatch will be executed on the GPU. For the specific requirements please refer to the HALCON “Installation Guide”.

'runtime_init'"runtime_init""runtime_init""runtime_init""runtime_init":

If called with 'immediately'"immediately""immediately""immediately""immediately", the GPU memory is initialized and the corresponding handle created. Otherwise this is done on demand, which may result in significantly larger execution times for the first call of apply_dl_classifierapply_dl_classifierApplyDlClassifierApplyDlClassifierApplyDlClassifier or train_dl_classifier_batchtrain_dl_classifier_batchTrainDlClassifierBatchTrainDlClassifierBatchTrainDlClassifierBatch. If 'gpu'"gpu""gpu""gpu""gpu" or 'batch_size_device'"batch_size_device""batch_size_device""batch_size_device""batch_size_device" is changed with subsequent calls of set_dl_classifier_paramset_dl_classifier_paramSetDlClassifierParamSetDlClassifierParamSetDlClassifierParam, the GPU memory is reinitialized.

Note, this parameter has no effect if running on CPUs, thus if 'runtime'"runtime""runtime""runtime""runtime" is set to 'cpu'"cpu""cpu""cpu""cpu".

'weight_prior'"weight_prior""weight_prior""weight_prior""weight_prior":

Regularization parameter used for regularization of the loss function. Regularization is helpful in the presence of overfitting during the classifier training. If the hyperparameter 'weight_prior'"weight_prior""weight_prior""weight_prior""weight_prior" is non-zero, the regularization term given below is added to the loss function (see also train_dl_classifier_batchtrain_dl_classifier_batchTrainDlClassifierBatchTrainDlClassifierBatchTrainDlClassifierBatch) Here the index k runs over all weights of the network, except for the biases which are not regularized. The regularization term generally penalizes large weights, thus pushing the weights towards zero, which effectively reduces the complexity of the model. Simply put: Regularization favors simpler models that are less likely to learn noise in the data and generalize better. Per default no regularization is used, i.e. 'weight_prior'"weight_prior""weight_prior""weight_prior""weight_prior" is set to 0.0. In case the classifier overfits the data, it is strongly recommended to try different values for the parameter 'weight_prior'"weight_prior""weight_prior""weight_prior""weight_prior" to improve the generalization properties of the neural network. Choosing its value is a trade-off between the models ability to generalize, overfitting, and underfitting. If is too small, the model might overfit, if its too large the model might loose its ability to fit the data, because all weights are effectively zero. For finding an ideal value for , we recommend a cross-validation, i.e. to perform the training for a range of values and choose the value that results in the best validation error. For typical applications, we recommend testing the values for 'weight_prior'"weight_prior""weight_prior""weight_prior""weight_prior" on a logarithmic scale between . If the training takes a very long time, one might consider performing the hyperparameter optimization on a reduced amount of data.

For an explanation of the concept of deep-learning-based classification see the introduction of chapter Deep Learning / Classification.

Attention

To successfully set 'gpu'"gpu""gpu""gpu""gpu" parameters, cuDNN and cuBLAS are required, i.e., to set the parameter GenParamNameGenParamNameGenParamNameGenParamNamegenParamName 'runtime'"runtime""runtime""runtime""runtime" to VarRef('gpu') or to set the GenParamNameGenParamNameGenParamNameGenParamNamegenParamName 'gpu'"gpu""gpu""gpu""gpu". For further details, please refer to the “Installation Guide”, paragraph “Requirements for Deep Learning”.

Execution Information

Parameters

DLClassifierHandleDLClassifierHandleDLClassifierHandleDLClassifierHandleDLClassifierHandle (input_control)  dl_classifier HDlClassifier, HTupleHTupleHtuple (handle) (IntPtr) (HHandle) (handle)

Handle of the deep-learning-based classifier.

GenParamNameGenParamNameGenParamNameGenParamNamegenParamName (input_control)  attribute.name(-array) HTupleHTupleHtuple (string) (string) (HString) (char*)

Name of the generic parameter.

Default value: 'classes' "classes" "classes" "classes" "classes"

List of values: 'batch_size'"batch_size""batch_size""batch_size""batch_size", 'batch_size_device'"batch_size_device""batch_size_device""batch_size_device""batch_size_device", 'classes'"classes""classes""classes""classes", 'gpu'"gpu""gpu""gpu""gpu", 'image_dimensions'"image_dimensions""image_dimensions""image_dimensions""image_dimensions", 'image_height'"image_height""image_height""image_height""image_height", 'image_num_channels'"image_num_channels""image_num_channels""image_num_channels""image_num_channels", 'image_width'"image_width""image_width""image_width""image_width", 'learning_rate'"learning_rate""learning_rate""learning_rate""learning_rate", 'momentum'"momentum""momentum""momentum""momentum", 'runtime'"runtime""runtime""runtime""runtime", 'runtime_init'"runtime_init""runtime_init""runtime_init""runtime_init", 'weight_prior'"weight_prior""weight_prior""weight_prior""weight_prior"

GenParamValueGenParamValueGenParamValueGenParamValuegenParamValue (input_control)  attribute.value(-array) HTupleHTupleHtuple (string / real / integer) (string / double / int / long) (HString / double / Hlong) (char* / double / Hlong)

Value of the generic parameter.

Default value: ['class_1','class_2','class_3'] ["class_1","class_2","class_3"] ["class_1","class_2","class_3"] ["class_1","class_2","class_3"] ["class_1","class_2","class_3"]

Suggested values: 1, 2, 3, 50, 0.001, 'cpu'"cpu""cpu""cpu""cpu", 'gpu'"gpu""gpu""gpu""gpu", 'immediately'"immediately""immediately""immediately""immediately"

Result

If the parameters are valid, the operator set_dl_classifier_paramset_dl_classifier_paramSetDlClassifierParamSetDlClassifierParamSetDlClassifierParam returns the value 2 (H_MSG_TRUE). If necessary, an exception is raised.

Possible Predecessors

read_dl_classifierread_dl_classifierReadDlClassifierReadDlClassifierReadDlClassifier

Possible Successors

get_dl_classifier_paramget_dl_classifier_paramGetDlClassifierParamGetDlClassifierParamGetDlClassifierParam, apply_dl_classifierapply_dl_classifierApplyDlClassifierApplyDlClassifierApplyDlClassifier, train_dl_classifier_batchtrain_dl_classifier_batchTrainDlClassifierBatchTrainDlClassifierBatchTrainDlClassifierBatch

See also

get_dl_classifier_paramget_dl_classifier_paramGetDlClassifierParamGetDlClassifierParamGetDlClassifierParam

Module

Deep Learning Inference