apply_deep_counting_model — Apply a Deep Counting model on a set of images for inference.
apply_deep_counting_model(Image : : DeepCountingHandle : Count, DeepCountingResult)
apply_deep_counting_model applies the Deep Counting model given by
DeepCountingHandle on the tuple of input images Image.
A tuple with the number of found instances per input image is
returned in Count.
Additional information about the counted instances is returned in
DeepCountingResult, which contains a tuple of dictionaries,
again one per input image.
Note that templates of the objects to be counted must be set
beforehand using prepare_deep_counting_model.
Note that the inference time of apply_deep_counting_model depends
on the number of defined templates in prepare_deep_counting_model.
Please see the chapter Matching / Deep Counting for further information.
The procedure dev_display_deep_count_results can be used to
visualize the detected instances.
The settings 'min_score' and 'max_overlap'
can be used to set the minimum similarity of the instances
to the templates as well as the allowed overlap of instances.
Details can be found in get_deep_counting_model_param.
The dictionary returned in DeepCountingResult contains
additional information about the detected and counted object instances.
It contains the following keys, where each key contains a tuple of values,
one for each detection. If no instances were detected, i.e. the count
is 0, those tuples are empty.
Approximate area of the detected templates in pixels.
Approximate location in row and column coordinates of the instances in the input image.
Similarity score of the detected instances, i.e., their approximate
similarity with the most similar template provided to
prepare_deep_counting_model.
Index of the template with the highest similarity to the detected instance. This can be used to find out which templates were counted.
The angle and scale of the template with the highest similarity
to the detected instance.
If rotational or scale augmentations were enabled during the call of
prepare_deep_counting_model, these values can be used
to find out which rotation angle and scale creates a template that is
most similar to the detected instance.
Deep Counting does not take into account whether an image has a reduced domain.
System requirements:
To run this operator on GPU (see get_deep_counting_model_param),
cuDNN and cuBLAS are required.
For further details, please refer to the “Installation Guide”,
paragraph “Requirements for Deep Learning and Deep-Learning-Based Methods”.
Alternatively, this operator can also be run on CPU.
This operator returns a handle. Note that the state of an instance of this handle type may be changed by specific operators even though the handle is used as an input parameter by those operators.
Image (input_object) (multichannel-)image(-array) → object (byte / real)
Input image.
DeepCountingHandle (input_control) deep_counting → (handle)
Handle of the Deep Counting model.
Count (output_control) integer(-array) → (integer)
Number of counted objects.
DeepCountingResult (output_control) dict(-array) → (handle)
Tuple of result dictionaries.
If the parameters are valid, the operator apply_deep_counting_model
returns the value 2 (
H_MSG_TRUE)
. If necessary, an exception is raised.
create_deep_counting_model,
set_deep_counting_model_param,
get_deep_counting_model_param,
prepare_deep_counting_model
Matching