OpenVX Classifier Extension  9249ea0
Scan Classifier

Scans a feature-map (input_feature_map) and do the classification for each scan-window. More...

Functions

vx_node vxScanClassifierNode (vx_graph graph, vx_tensor input_feature_map, vx_classifier_model model, vx_int32 scanwindow_width, vx_int32 scanwindow_height, vx_int32 step_x, vx_int32 step_y, vx_array object_confidences, vx_array object_rectangles, vx_scalar num_objects)
 [Graph] Scans a feature-map (input_feature_map) and detect the classification for each scan-window. More...
 

Detailed Description

Scans a feature-map (input_feature_map) and do the classification for each scan-window.

This function scans a feature-map. Each window in the feature map is classified by a classification model. The classification models are loadable by undefined binary format see vxImportClassifierModel. Extensions will be added to the specification, to support a defined binary format. Classification models can be any machine learning classification method. Examples are Cascade, SVM, and Neural Networks.

Function Documentation

§ vxScanClassifierNode()

vx_node vxScanClassifierNode ( vx_graph  graph,
vx_tensor  input_feature_map,
vx_classifier_model  model,
vx_int32  scanwindow_width,
vx_int32  scanwindow_height,
vx_int32  step_x,
vx_int32  step_y,
vx_array  object_confidences,
vx_array  object_rectangles,
vx_scalar  num_objects 
)

[Graph] Scans a feature-map (input_feature_map) and detect the classification for each scan-window.

Parameters
[in]graphThe reference to the graph
[in]input_feature_mapThe Feature-map, example is the output of vxHOGFeaturesNode.
[in]modelThe pre-trained model loaded. Loaded using vxImportClassifierModel
[in]scan_window_widthWidth of the scan window
[in]scan_window_heightHeight of the scan window
[in]step_xHorizontal step-size (along x-axis)
[in]step_yVertical step-size (along y-axis)
[out]object_confidences[Optional] An array of confidences measure, the measure is of type VX_TYPE_UINT16. The confidence measure is defined by the extensions which define classification model with defined binary format. This output can be used as class index as well. In case we detect several different classes in single execution. The output will be an array of indexes of the classes.
[out]object_rectanglesAn array of object positions, in VX_TYPE_RECTANGLE
[out]num_objects[optional] The number of object detected in a VX_SIZE scalar
Note
The border mode VX_NODE_BORDER value VX_BORDER_UNDEFINED is supported.
Returns
vx_node.
Return values
vx_nodeA node reference. Any possible errors preventing a successful creation should be checked using vxGetStatus