get_deep_counting_model_param🔗
Short description🔗
get_deep_counting_model_param — Return the parameters of a Deep Counting model.
Signature🔗
get_deep_counting_model_param( deep_counting DeepCountingHandle, attribute.name GenParamName, out attribute.name GenParamValue )
Description🔗
The operator get_deep_counting_model_param returns the parameter values
of GenParamName for the Deep Counting model DeepCountingHandle
in GenParamValue.
Note that when changing parameters that influence the template creation,
prepare_deep_counting_model must be called again before
the model can be applied with apply_deep_counting_model.
The following table gives an overview,
which parameters can be set
using set_deep_counting_model_param or create_deep_counting_model
(set),
which can be retrieved using get_deep_counting_model_param
(get),
and which require re-running prepare_deep_counting_model after
changing them (prepare).
GenParamName |
set |
get |
Requires prepare |
|---|---|---|---|
| 'angle_start' | x |
x |
x |
| 'angle_step' | x |
x |
x |
| 'angle_end' | x |
x |
x |
| 'backbone_model' | x |
x |
x |
| 'device' | x |
x |
x |
| 'max_overlap' | x |
x |
|
| 'min_score' | x |
x |
|
| 'scale_max' | x |
x |
x |
| 'scale_min' | x |
x |
x |
| 'scale_step' | x |
x |
x |
In the following the parameters are described:
-
'angle_start', 'angle_step', 'angle_end': Control the rotational augmentation. Templates passed to
prepare_deep_counting_modelare rotated from 'angle_start' to 'angle_end' in steps of 'angle_step'. This allowsapply_deep_counting_modelto better find rotated instances of the templates.The angles must be passed in radians.
Suggested values: 0, -6.28, -3.14, 3.14, 6.28
Default: 'angle_start' = 0, 'angle_end' = 0, 'angle_step' = 'rad(30)'
Restriction: \(-2\pi\) <= 'angle_start' <= 'angle_end' <= \(2\pi\), 'angle_step' > 0
-
'backbone_model': The backbone used for the detection of the templates. The backbone is automatically created by
create_deep_counting_model. It can be obtained and written back in order to, for example, optimize it usingoptimize_dl_model_for_inference.Note that the Deep Counting model will automatically set the input size of the backbone according to the template and image sizes. It has therefore no effect to change the backbone’s input size, and it is not recommended to do so. Also note that the backbone can not be used for any other deep learning methods besides Deep Counting.
-
'device': Handle of the device on which the backbone will be executed.
If the backbone was already optimized for a device, setting 'device' might not be necessary anymore, see
optimize_dl_model_for_inferencefor details.To get a tuple of handles of all available potentially deep-learning capable hardware devices use
query_available_dl_devices.Default: Handle of the default device, thus the GPU with index 0. If not available, this is an empty tuple.
-
'max_overlap': The maximum allowed intersection over union (IoU) for two detected templates during counting. When two templates have an IoU higher than 'max_overlap', the one with lower confidence value gets suppressed. When set to 0, no overlap at all is allowed. We refer to the chapter Deep Learning / Object Detection for further explanations of the IoU.
Suggested values: 0.3, 0.5, 0.7, 1.0
Default: 'max_overlap' = 0.5
Restriction: 0 <= 'max_overlap' <= 1
-
'min_score': This parameter determines the minimum similarity of detected instances to the original templates. In other words,
apply_deep_counting_modelignores all detected instances with a similarity smaller than this value. The similarity computed by the Deep Counting model lies between 0 and 1, where 0 means no similarity and 1 is a very high similarity.Suggested values: 0.2, 0.3, 0.4, 0.5
Default: 'min_score' = 0.4
Restriction: 0 < 'min_score' <= 1
-
'scale_min', 'scale_step', 'scale_max': Control the scale augmentation. Templates passed to
prepare_deep_counting_modelare scaled from 'scale_min' to 'scale_max' in steps of 'scale_step'. This allowsapply_deep_counting_modelto better find scaled instances of the templates.Suggested values: 0.9, 1.0, 1.1
Default: 'scale_min' = 1.0, 'scale_max' = 1.0, 'scale_step' = 0.1
Restriction: 0 < 'scale_min' <= 'scale_max', 'scale_step' > 0
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.
Parameters🔗
DeepCountingHandle (input_control) deep_counting → (handle)
Handle of the Deep Counting model.
GenParamName (input_control) attribute.name → (string)
Name of the generic parameter.
Default: 'angle_start'
List of values: 'angle_end', 'angle_start', 'angle_step', 'backbone_model', 'device', 'max_overlap', 'min_score', 'scale_max', 'scale_min', 'scale_step'
GenParamValue (output_control) attribute.name → (real / string / integer)
Value of the generic parameter.
Result🔗
If the handle of the model is valid, the operator
get_deep_counting_model_param returns the value 2 (H_MSG_TRUE). If necessary
an exception is raised.
Combinations with other operators🔗
Combinations
Possible predecessors
create_deep_counting_model, set_deep_counting_model_param, read_deep_counting_model
Possible successors
See also
Module🔗
Matching