create_dl_layer_grid_sample🔗
Short description🔗
create_dl_layer_grid_sample — Create a grid sample layer.
Signature🔗
create_dl_layer_grid_sample( dl_layer DLLayerInput, dl_layer DLLayerGrid, string LayerName, attribute.name GenParamName, attribute.value GenParamValue, out dl_layer DLLayerGridSample )
Description🔗
The operator create_dl_layer_grid_sample creates a grid sample layer,
which applies a spatial transformation to an input image or feature map.
The handle of the created grid sample layer is returned in
DLLayerGridSample.
A grid sample layer warps the input DLLayerInput based on a sampling
grid DLLayerGrid. This grid needs to specify spatially normalized
input pixel locations which are then used to interpolate the output values.
Pixel locations inside the input image or feature map must have values in
the range of \([-1, 1]\).
Thus, the values \(x=-1\), \(y=-1\) must correspond to the
top-left pixel, while \(x=1\), \(y=1\) describe the bottom-right
pixel. Thereby, the extrema refer to the center points of the corner pixels.
The grid sample layer internally sets ‘align_corners’ to 1, defining
the pixel centers as the grid reference points. The sampling grid
DLLayerGrid can be generated using create_dl_layer_affine_grid.
The grid sample layer uses bilinear interpolation to compute output values, ensuring smooth transitions between pixel values. For sampling grid locations outside of the input image or feature map the corresponding values are set to 0. The output of the grid sample layer has the same spatial dimensions as the input grid.
The parameter LayerName sets an individual layer name.
Note that if creating a model using create_dl_model each layer of
the created network must have a unique name.
The following generic parameters GenParamName and the corresponding
values GenParamValue are supported:
-
'is_inference_output': Determines whether
apply_dl_modelwill include the output of this layer in the dictionaryDLResultBatcheven without specifying this layer inOutputs('true') or not ('false').List of values: 'true', 'false'
Default: 'false'
Certain parameters of layers created using this operator
create_dl_layer_grid_sample can be set and retrieved using
further operators.
The following tables give an overview, which parameters can be set
using set_dl_model_layer_param and which ones can be retrieved
using get_dl_model_layer_param or get_dl_layer_param. Note, the
operators set_dl_model_layer_param and get_dl_model_layer_param
require a model created by create_dl_model.
| Layer Parameters | set |
get |
|---|---|---|
'input_layer' (DLLayerInput, DLLayerGrid) |
x |
|
'name' (LayerName) |
x |
x |
'output_layer' (DLLayerGridSample) |
x |
|
| 'shape' | x |
|
| 'type' | x |
| Generic Layer Parameters | set |
get |
|---|---|---|
| 'is_inference_output' | x |
x |
| 'num_trainable_params' | x |
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🔗
DLLayerInput (input_control) dl_layer → (handle)
Input layer for image or feature map.
DLLayerGrid (input_control) dl_layer → (handle)
Input layer defining the sampling grid.
LayerName (input_control) string → (string)
Name of the output layer.
GenParamName (input_control) attribute.name(-array) → (string)
Generic input parameter names.
Default: []
List of values: 'is_inference_output'
GenParamValue (input_control) attribute.value(-array) → (string / integer / real)
Generic input parameter values.
Default: []
Suggested values: 'true', 'false'
DLLayerGridSample (output_control) dl_layer → (handle)
Grid sample layer.
Example🔗
(HDevelop)
read_image (Image, 'printer_chip/printer_chip_01')
convert_image_type (Image, Image, 'real')
get_image_size (Image, Width, Height)
* Create input layer for the image to be transformed.
create_dl_layer_input ('image', [Width,Height,1], [], [], DLLayerInputImage)
*
* Create an affine grid layer and define the size of its output grid with (Width/2, Height/2), which later
* determines the output image size.
create_dl_layer_input ('transformation', [3,2,1], [], [], DLLayerInputTransformation)
create_dl_layer_affine_grid (DLLayerInputTransformation, 'affine_grid', Width/2, Height/2, [], [], DLLayerAffineGrid)
*
* Create an affine transformation which represents a translation
* to the right by half the image width.
* As the affine grid layer uses normalized pixel positions in
* the range of [-1,1] the translation value must be normalized
* as well. Hence, translating by half the image width to the right
* corresponds to a translation value of -1 in x-direction.
hom_mat2d_identity (HomMat2DIdentity)
hom_mat2d_translate (HomMat2DIdentity, -1, 0, HomMat2DTranslate)
*
* Create grid sample layer with the affine grid as input
create_dl_layer_grid_sample (DLLayerInputImage, DLLayerAffineGrid, 'output', [], [], DLLayerGridSample)
*
* Create and apply dl model
create_dl_model (DLLayerGridSample, DLModelHandle)
DLSample := dict{image: Image, transformation: HomMat2DTranslate}
apply_dl_model (DLModelHandle, DLSample, [], DLResultBatch)
get_dict_object (ImageTranslated, DLResultBatch, 'output')
Combinations with other operators🔗
Module🔗
Deep Learning Professional