AMD WinML Extension API

Functions

Extension: AMD WinML Extension API

AMD OpenVX WinML Interop Nodes Extension [only supported on Windows]. More...

Functions
SHARED_PUBLIC vx_node VX_API_CALL	vxExtWinMLNode_OnnxToMivisionX (vx_graph graph, vx_scalar modelLocation, vx_scalar inputTensorName, vx_scalar outputTensorName, vx_tensor inputTensor, vx_array setupArray, vx_tensor outputTensor, vx_scalar deviceKind)
	[Graph] Creates a WinML import ONNX Model and run function node. More...
SHARED_PUBLIC vx_node VX_API_CALL	vxExtWinMLNode_convertImageToTensor (vx_graph graph, vx_image input, vx_tensor output, vx_scalar a, vx_scalar b, vx_scalar reverse_channel_order)
	[Graph] Creates a WinML convert image to tensor node. More...
SHARED_PUBLIC vx_node VX_API_CALL	vxExtWinMLNode_getTopKLabels (vx_graph graph, vx_tensor prob_tensor, vx_scalar labelFile, vx_scalar output_1, vx_scalar output_2, vx_scalar output_3, vx_scalar output_4, vx_scalar output_5)
	[Graph] Creates a output tensor to Top K labels node. More...

Detailed Description

AMD OpenVX WinML Interop Nodes Extension [only supported on Windows].

Function Documentation

◆ vxExtWinMLNode_convertImageToTensor()

SHARED_PUBLIC vx_node VX_API_CALL vxExtWinMLNode_convertImageToTensor	(	vx_graph	graph,
		vx_image	input,
		vx_tensor	output,
		vx_scalar	a,
		vx_scalar	b,
		vx_scalar	reverse_channel_order
	)

[Graph] Creates a WinML convert image to tensor node.

Note

Kernel Name: com.winml.convert_image_to_tensor

Parameters

[in]	graph	The reference to the graph.
[in]	input_1	input image vx_image.
[out]	The	output Tensor in VX_FLOAT32 format.
[in]	input_2	a in vx_scalar.
[in]	input_3	b in vx_scalar.
[in]	input_4	reverse channel order in vx_scalar.

Returns

vx_node.

Return values

vx_node

A node reference. Any possible errors preventing a successful creation should be checked using vxGetStatus

◆ vxExtWinMLNode_getTopKLabels()

SHARED_PUBLIC vx_node VX_API_CALL vxExtWinMLNode_getTopKLabels	(	vx_graph	graph,
		vx_tensor	prob_tensor,
		vx_scalar	labelFile,
		vx_scalar	output_1,
		vx_scalar	output_2,
		vx_scalar	output_3,
		vx_scalar	output_4,
		vx_scalar	output_5
	)

[Graph] Creates a output tensor to Top K labels node.

Note

Kernel Name: com.winml.get_top_k_labels

Parameters

[in]	graph	The reference to the graph.
[in]	input_1	probability tensor vx_tensor.
[in]	input_2	label text file location vx_scalar.
[out]	output_1	Top 1 in vx_scalar.
[out]	output_2	Top 2 in vx_scalar. [optional]
[out]	output_3	Top 3 in vx_scalar. [optional]
[out]	output_4	Top 4 in vx_scalar. [optional]
[out]	output_5	Top 5 in vx_scalar. [optional]

Returns

vx_node.

Return values

vx_node

A node reference. Any possible errors preventing a successful creation should be checked using vxGetStatus

◆ vxExtWinMLNode_OnnxToMivisionX()

SHARED_PUBLIC vx_node VX_API_CALL vxExtWinMLNode_OnnxToMivisionX	(	vx_graph	graph,
		vx_scalar	modelLocation,
		vx_scalar	inputTensorName,
		vx_scalar	outputTensorName,
		vx_tensor	inputTensor,
		vx_array	setupArray,
		vx_tensor	outputTensor,
		vx_scalar	deviceKind
	)

[Graph] Creates a WinML import ONNX Model and run function node.

---

WinML VX_API_ENTRY C Function NODE

Note

Kernel Name: com.winml.onnx_to_mivisionx

Parameters

[in]	graph	The reference to the graph.
[in]	input_1	The ONNX Model Location in vx_scalar.
[in]	input_2	The ONNX Model Input Tensor Name in vx_scalar.
[in]	input_3	The ONNX Model Output Tensor Name in vx_scalar.
[in]	input_4	The Input Tensor in VX_FLOAT32 format.
[in]	input_5	The setup Array for each model in VX_TYPE_SIZE format.
[out]	output	The output Tensor in VX_FLOAT32 format.
[in]	input_6	WinML Deploy Device Kind in vx_scalar [optional] (default: 0).

Returns

vx_node.

Return values

vx_node

A node reference. Any possible errors preventing a successful creation should be checked using vxGetStatus