Introduction

This section documents the kernels defined for Image Pre/Post Processing.

Data Structures
struct	tivxImgPreProcParams

struct	tivxOCPreProcParams

struct	tivxOCPostProcParams

struct	tivxOCPostProcOutput

struct	tivxODPostProcParams

struct	tivxDofPlaneSepParams

struct	tivxVisualLocalizationParams

struct	tivxPixelVizParams

struct	tivxImgMosaicParams

struct	tivxDrawKeypointDetectionsParams

struct	tivx3DODPostProcParams

struct	tivxDrawBoxDetectionsParams

struct	tivxDraw3DBoxDetectionsParams

struct	tivxSFMParams
	SFM frame level control parameter. More...

struct	tivxDLPreProcParams
	DL Pre processing to be used with DL-RT. More...

struct	tivxDLColorBlendParams
	DL color blend to be used with DL-RT. More...

struct	tivxDLRectangle
	DL rectangles to be used with DL Draw Box. More...

struct	tivxDLDrawBoxParams
	DL draw box to be used with DL-RT. More...

Functions
void	tivxImgProcLoadKernels (vx_context context)
	Used for the Application to load the img_proc kernels into the context. More...

void	tivxImgProcUnLoadKernels (vx_context context)
	Used for the Application to unload the img_proc kernels from the context. More...

vx_kernel	tivxAddKernelPixelViz (vx_context context, vx_int32 num_output_tensors)
	Used for the Application to create the pix viz kernel from the context. More...

vx_kernel	tivxAddKernelImgMosaic (vx_context context, vx_int32 num_inputs)
	Used by the application to create the mosaic kernel from the context. More...

void	tivxImgMosaicParamsSetDefaults (tivxImgMosaicParams *prms)
	Set default parameters for tivxImgMosaicParams. More...

VX_API_ENTRY vx_node VX_API_CALL	tivxImgPreProcNode (vx_graph graph, vx_array config, vx_image in_img, vx_tensor out_tensor)
	Creates a Image Pre Processing Node. More...

VX_API_ENTRY vx_node VX_API_CALL	tivxOCPreProcNode (vx_graph graph, vx_user_data_object config, vx_image in_img, vx_tensor out_tensor)
	Creates a OC Pre Processing Node. More...

VX_API_ENTRY vx_node VX_API_CALL	tivxOCPostProcNode (vx_graph graph, vx_user_data_object config, vx_user_data_object in_args, vx_tensor in_tensor, vx_user_data_object results)
	Creates a OC Post Processing Node. More...

VX_API_ENTRY vx_node VX_API_CALL	tivxODPostProcNode (vx_graph graph, vx_array configuration, vx_tensor input_key_points, vx_tensor fwd_table_tensor, vx_tensor rev_table_tensor, vx_tensor kp_tensor, vx_tensor kp_valid)
	Creates a Post processing Node for Parking Spot Detection Algorithm. More...

VX_API_ENTRY vx_node VX_API_CALL	tivxDofPlaneSepNode (vx_graph graph, vx_user_data_object config, vx_image dof_flow, vx_tensor out_planes)
	Creates a Dense Optical Flow (Dof) 3 plane creation Node. More...

VX_API_ENTRY vx_node VX_API_CALL	tivxPixelVizNode (vx_graph graph, vx_kernel kernel, vx_user_data_object configuration, vx_user_data_object tidl_out_args, vx_image in_img, vx_int32 num_output_tensors, vx_tensor detection[], vx_image out_img[])
	Creates an image with pixel level detection/label overlayed. More...

VX_API_ENTRY vx_node VX_API_CALL	tivxVisualLocalizationNode (vx_graph graph, vx_array configuration, vx_tensor voxel_info, vx_tensor map_3d_points, vx_tensor map_desc, vx_tensor ext_feat_pt, vx_tensor ext_feat_desc, vx_tensor up_samp_wt, vx_tensor lens_dist_table, vx_user_data_object tidl_out_args, vx_matrix pose_tensor)
	Estimating the camera location from captured image and map information. More...

VX_API_ENTRY vx_node VX_API_CALL	tivxPoseVizNode (vx_graph graph, vx_array configuration, vx_image background_image, vx_matrix pose_tensor, vx_image output_image)
	Visualization of the estimated location. More...

VX_API_ENTRY vx_node VX_API_CALL	tivxImgMosaicNode (vx_graph graph, vx_kernel kernel, vx_user_data_object config, vx_image output_image, vx_image background_image, vx_object_array input_arr[], vx_uint32 num_inputs)
	Image Mosiac Node. More...

VX_API_ENTRY vx_node VX_API_CALL	tivxDrawKeypointDetectionsNode (vx_graph graph, vx_user_data_object configuration, vx_tensor kp_tensor, vx_tensor kp_valid, vx_tensor input_tensor, vx_image input_image, vx_image output_image)
	Draw Keypoint Detections Node. More...

VX_API_ENTRY vx_node VX_API_CALL	tivxDrawBoxDetectionsNode (vx_graph graph, vx_user_data_object configuration, vx_tensor input_tensor, vx_image input_image, vx_image output_image)
	Draw Box Detections Node. More...

VX_API_ENTRY vx_node VX_API_CALL	tivxDLBEVPostProcNode (vx_graph graph, vx_kernel kernel, vx_user_data_object config, vx_image input_image, vx_tensor Lidar_2_Img_tensor, vx_tensor input_tensor[], vx_image output_image)
	Draw BEV Post Proc Node. More...

VX_API_ENTRY vx_node VX_API_CALL	tivxImgHistNode (vx_graph graph, vx_image input_image, vx_distribution output_histogram)
	Image Histogram Node. More...

VX_API_ENTRY vx_node VX_API_CALL	tivxSFMNode (vx_graph graph, vx_user_data_object config, vx_user_data_object create_params, vx_user_data_object in_args, vx_user_data_object out_args, vx_image input_image, vx_image flow_vectors, vx_image output_ptcld_image, vx_image output_og_image, vx_user_data_object output_feat)
	Structure From Motion Node. More...

VX_API_ENTRY vx_node VX_API_CALL	tivxDLPreProcNode (vx_graph graph, vx_user_data_object config, vx_image input_image, vx_tensor output_tensor)
	DL Pre-processing Node. More...

VX_API_ENTRY vx_node VX_API_CALL	tivxDLColorBlendNode (vx_graph graph, vx_user_data_object config, vx_image input_image, vx_tensor input_tensor, vx_image output_image)
	DL Color Blend Node. More...

VX_API_ENTRY vx_node VX_API_CALL	tivxDLDrawBoxNode (vx_graph graph, vx_user_data_object config, vx_image input_image, vx_tensor input_tensor, vx_image output_image)
	DL Draw Box Node. More...

VX_API_ENTRY vx_node VX_API_CALL	tivxDLColorConvertNode (vx_graph graph, vx_image input_image, vx_image output_image)
	DL Color Convert Node. More...

Macros
#define	TIVX_MODULE_NAME_IMG_PROC "img_proc"
	OpenVX module name. More...

#define	TIVX_PIXEL_VIZ_MAX_TENSOR (3U)

Macro Definition Documentation

◆ TIVX_MODULE_NAME_IMG_PROC

#define TIVX_MODULE_NAME_IMG_PROC "img_proc"

OpenVX module name.

◆ TIVX_PIXEL_VIZ_MAX_TENSOR

#define TIVX_PIXEL_VIZ_MAX_TENSOR (3U)

Function Documentation

◆ tivxImgProcLoadKernels()

void tivxImgProcLoadKernels ( vx_context context )

Used for the Application to load the img_proc kernels into the context.

◆ tivxImgProcUnLoadKernels()

void tivxImgProcUnLoadKernels ( vx_context context )

Used for the Application to unload the img_proc kernels from the context.

◆ tivxAddKernelPixelViz()

vx_kernel tivxAddKernelPixelViz	(	vx_context	context,
		vx_int32	num_output_tensors
	)

Used for the Application to create the pix viz kernel from the context.

◆ tivxAddKernelImgMosaic()

vx_kernel tivxAddKernelImgMosaic	(	vx_context	context,
		vx_int32	num_inputs
	)

Used by the application to create the mosaic kernel from the context.

◆ tivxImgMosaicParamsSetDefaults()

void tivxImgMosaicParamsSetDefaults ( tivxImgMosaicParams * prms )

Set default parameters for tivxImgMosaicParams.

◆ tivxImgPreProcNode()

VX_API_ENTRY vx_node VX_API_CALL tivxImgPreProcNode	(	vx_graph	graph,
		vx_array	config,
		vx_image	in_img,
		vx_tensor	out_tensor
	)

Creates a Image Pre Processing Node.

Node does pre processing needed for conventional deep learning algorithm. Preprocessing steps such as mean subtraction, scaling and plane deinterleaving is supported by this Node.

Parameters

[in]	graph	The reference to the graph.
[in]	config	Array element type tivxImgPreProcParams, max number of elements = 1
[in]	in_img	input image
[in]	out_tensor	output tensor

Return values

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

◆ tivxOCPreProcNode()

VX_API_ENTRY vx_node VX_API_CALL tivxOCPreProcNode	(	vx_graph	graph,
		vx_user_data_object	config,
		vx_image	in_img,
		vx_tensor	out_tensor
	)

Creates a OC Pre Processing Node.

Node does pre processing needed for image classification operation Performs only NV12 to BGR conversion

Parameters

[in]	graph	The reference to the graph.
[in]	config	User data object element type tivxOCPreProcParams
[in]	in_img	input image
[out]	out_tensor	output tensor

Return values

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

◆ tivxOCPostProcNode()

VX_API_ENTRY vx_node VX_API_CALL tivxOCPostProcNode	(	vx_graph	graph,
		vx_user_data_object	config,
		vx_user_data_object	in_args,
		vx_tensor	in_tensor,
		vx_user_data_object	results
	)

Creates a OC Post Processing Node.

Node does post processing needed for conventional image classification algorithm. Returns top N classes from a list of probabilities

Parameters

[in]	graph	The reference to the graph.
[in]	config	User data object element type tivxOCPostProcParams
[in]	in_args	input user data object
[in]	in_tensor	input tensor
[out]	results	output user data object

Return values

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

◆ tivxODPostProcNode()

VX_API_ENTRY vx_node VX_API_CALL tivxODPostProcNode	(	vx_graph	graph,
		vx_array	configuration,
		vx_tensor	input_key_points,
		vx_tensor	fwd_table_tensor,
		vx_tensor	rev_table_tensor,
		vx_tensor	kp_tensor,
		vx_tensor	kp_valid
	)

Creates a Post processing Node for Parking Spot Detection Algorithm.

Node does generates interpolated points on parking spot periphery from four corners points detected from deep learning based algorithm

Parameters

[in]	graph	The reference to the graph.
[in]	configuration	Array element type tivxODPostProcParams, max number of elements = 1
[in]	input_key_points	Interleaved 4 corner points of multiple parking spots.
[in]	fwd_table_tensor	Angular map table for rectilinear angle to fish eye angle
[in]	rev_table_tensor	Angular map table for fish eye angle to rectilinear angle
[in]	kp_tensor	output interpolated points on periphery of parking spot
[in]	kp_valid	output flag of validity for each parking spot

Return values

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

◆ tivxDofPlaneSepNode()

VX_API_ENTRY vx_node VX_API_CALL tivxDofPlaneSepNode	(	vx_graph	graph,
		vx_user_data_object	config,
		vx_image	dof_flow,
		vx_tensor	out_planes
	)

Creates a Dense Optical Flow (Dof) 3 plane creation Node.

Node does creation of three planes (each of 8 bit) from the raw 32 bit output from Dof hardware accelerator 3 planes corresponds to horizontal, vertical flow vectors and confidence

Parameters

[in]	graph	The reference to the graph.
[in]	config	Array element type tivxDofPlaneSepParams, max number of elements = 1
[in]	dof_flow	input dof flow vectors
[out]	out_planes	output planes.

Return values

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

◆ tivxPixelVizNode()

VX_API_ENTRY vx_node VX_API_CALL tivxPixelVizNode	(	vx_graph	graph,
		vx_kernel	kernel,
		vx_user_data_object	configuration,
		vx_user_data_object	tidl_out_args,
		vx_image	in_img,
		vx_int32	num_output_tensors,
		vx_tensor	detection[],
		vx_image	out_img[]
	)

Creates an image with pixel level detection/label overlayed.

Node does blending of input image with pixel level detection/classification data for visualization purpose.

Parameters

[in]	graph	The reference to the graph.
[in]	kernel	The reference to openvx kernel
[in]	configuration	Array element type tivxODPostProcParams, max number of elements = 1
[in]	tidl_out_args	out args of tidl to receive the scale values of each output tensor
[in]	in_img	Input image.
[in]	num_output_tensors	Number of output tensors/images.
[in]	detection	Detection planes
[out]	out_img	ovelayed output image

Return values

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

◆ tivxVisualLocalizationNode()

VX_API_ENTRY vx_node VX_API_CALL tivxVisualLocalizationNode	(	vx_graph	graph,
		vx_array	configuration,
		vx_tensor	voxel_info,
		vx_tensor	map_3d_points,
		vx_tensor	map_desc,
		vx_tensor	ext_feat_pt,
		vx_tensor	ext_feat_desc,
		vx_tensor	up_samp_wt,
		vx_tensor	lens_dist_table,
		vx_user_data_object	tidl_out_args,
		vx_matrix	pose_tensor
	)

Estimating the camera location from captured image and map information.

Node estimates the location in 3d world co-ordinate

Parameters

[in]	graph	The reference to the graph.
[in]	configuration	Array element type tivxODPostProcParams, max number of elements = 1
[in]	voxel_info	Voxel info. voxels are cuboid in real world
[in]	map_3d_points	Map 3D points
[in]	map_desc	Map descriptors
[in]	ext_feat_pt	Externally computed feature points(x,y)
[in]	ext_feat_desc	Feature descriptor for each feature point
[in]	up_samp_wt	Filter weights for upsampling
[in]	lens_dist_table	Fish eye lens distortion table
[in]	tidl_out_args	TIDL outatgs needed for scale
[out]	pose_tensor	Output pose matrix of size 3x4

Return values

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

◆ tivxPoseVizNode()

VX_API_ENTRY vx_node VX_API_CALL tivxPoseVizNode	(	vx_graph	graph,
		vx_array	configuration,
		vx_image	background_image,
		vx_matrix	pose_tensor,
		vx_image	output_image
	)

Visualization of the estimated location.

Node generated an image with visualization of X, Y location of 3D world on image plane

Parameters

[in]	graph	The reference to the graph.
[in]	configuration	Array element type tivxODPostProcParams, max number of elements = 1
[in]	background_image	Background image on which pose information will be visualized
[in]	pose_tensor	Input pose matrix created by tivxVisualLocalizationNode Node
[out]	output_image	Output image

Return values

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

◆ tivxImgMosaicNode()

VX_API_ENTRY vx_node VX_API_CALL tivxImgMosaicNode	(	vx_graph	graph,
		vx_kernel	kernel,
		vx_user_data_object	config,
		vx_image	output_image,
		vx_image	background_image,
		vx_object_array	input_arr[],
		vx_uint32	num_inputs
	)

Image Mosiac Node.

Used to mosiac a list of images onto to a sigle image

Parameters

[in]	graph	The reference to the graph.
[in]	kernel	The reference to the openvx kernel
[in]	config	Mosaic configuration of type tivxImgMosaicParams
[out]	output_image	Output image of type vx_image, only VX_DF_IMAGE_NV12, VX_DF_IMAGE_U8, VX_DF_IMAGE_U16 supported
[in]	background_image	Optional background image of type vx_image, only VX_DF_IMAGE_NV12, VX_DF_IMAGE_U8, VX_DF_IMAGE_U16 supported
[in]	input_arr	Array of inputs of type vx_image, only VX_DF_IMAGE_NV12, VX_DF_IMAGE_U8, VX_DF_IMAGE_U16 supported
[in]	num_inputs	Number of inputs, i.e number of elements in input_arr[]

Return values

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

◆ tivxDrawKeypointDetectionsNode()

VX_API_ENTRY vx_node VX_API_CALL tivxDrawKeypointDetectionsNode	(	vx_graph	graph,
		vx_user_data_object	configuration,
		vx_tensor	kp_tensor,
		vx_tensor	kp_valid,
		vx_tensor	input_tensor,
		vx_image	input_image,
		vx_image	output_image
	)

Draw Keypoint Detections Node.

Used to draw keypoint detections on an image

Parameters

[in]	graph	reference to the graph.
[in]	configuration	configuration of type tivxDrawKeyPointDetectionParams
[in]	kp_tensor	keypoint tensor from OD postprocessing node
[in]	kp_valid	valid keypoints tensor from OD postprocessing node
[in]	input_tensor	TIDL output tensor of detected key points
[in]	input_image	Input image to draw detections
[out]	output_image	Output image with detections drawn

Return values

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

◆ tivxDrawBoxDetectionsNode()

VX_API_ENTRY vx_node VX_API_CALL tivxDrawBoxDetectionsNode	(	vx_graph	graph,
		vx_user_data_object	configuration,
		vx_tensor	input_tensor,
		vx_image	input_image,
		vx_image	output_image
	)

Draw Box Detections Node.

Used to draw box detections on an image

Parameters

[in]	graph	reference to the graph.
[in]	configuration	configuration of type tivxDrawBoxDetectionParams
[in]	input_tensor	TIDL output tensor of detected key points
[in]	input_image	Input image to draw detections
[out]	output_image	Output image with detections drawn

Return values

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

◆ tivxDLBEVPostProcNode()

VX_API_ENTRY vx_node VX_API_CALL tivxDLBEVPostProcNode	(	vx_graph	graph,
		vx_kernel	kernel,
		vx_user_data_object	config,
		vx_image	input_image,
		vx_tensor	Lidar_2_Img_tensor,
		vx_tensor	input_tensor[],
		vx_image	output_image
	)

Draw BEV Post Proc Node.

Used to draw box detections on an image

Parameters

[in]	graph	reference to the graph.
[in]	configuration	configuration of type tivxDrawBoxDetectionParams
[in]	input_tensor	TIDL output tensor of detected key points
[in]	input_image	Input image to draw detections
[out]	output_image	Output image with detections drawn

Return values

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

◆ tivxImgHistNode()

VX_API_ENTRY vx_node VX_API_CALL tivxImgHistNode	(	vx_graph	graph,
		vx_image	input_image,
		vx_distribution	output_histogram
	)

Image Histogram Node.

Used to get histogram of image intensity, done on A72

Parameters

[in]	graph	reference to the graph.
[in]	input_image	Input image to draw detections
[out]	output_histogram	Output histogram of image intensities (luma only)

Return values

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

◆ tivxSFMNode()

VX_API_ENTRY vx_node VX_API_CALL tivxSFMNode	(	vx_graph	graph,
		vx_user_data_object	config,
		vx_user_data_object	create_params,
		vx_user_data_object	in_args,
		vx_user_data_object	out_args,
		vx_image	input_image,
		vx_image	flow_vectors,
		vx_image	output_ptcld_image,
		vx_image	output_og_image,
		vx_user_data_object	output_feat
	)

Structure From Motion Node.

Structure From Motion (SFM), done on C71

Parameters

[in]	graph	reference to the graph.
[in]	config	SFM parameters defined in tivxSFMParams
[in]	create_params	SFM parameters defined in tivxSFMParams
[in]	in_args	SFM parameters defined in tivxSFMParams
[in]	out_args	SFM parameters defined in tivxSFMParams
[in]	input_image	Input image in NV12 format to overlay depth estimations
[in]	flow_vectors	Flow vectors in VX_DF_IMAGE_U32 format
[out]	output_ptcld_image	Output point cloud image
[out]	output_og_image	Output occupancy grid map
[out]	output_feat	Output features

Return values

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

◆ tivxDLPreProcNode()

VX_API_ENTRY vx_node VX_API_CALL tivxDLPreProcNode	(	vx_graph	graph,
		vx_user_data_object	config,
		vx_image	input_image,
		vx_tensor	output_tensor
	)

DL Pre-processing Node.

Transforms input image to a format as required for DL processing

Parameters

[in]	graph	reference to the graph
[in]	config	reference to tivxDLPreProcPrams
[in]	input_image	Input image
[out]	output_tensor	Output tensor

Return values

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

◆ tivxDLColorBlendNode()

VX_API_ENTRY vx_node VX_API_CALL tivxDLColorBlendNode	(	vx_graph	graph,
		vx_user_data_object	config,
		vx_image	input_image,
		vx_tensor	input_tensor,
		vx_image	output_image
	)

DL Color Blend Node.

Takes the DL output and blends the provided color-map to the input image

Parameters

[in]	graph	reference to the graph
[in]	config	reference to tivxDLPreProcPrams
[in]	input_image	Input image
[in]	input_tensor	Input tensor
[out]	output_image	Output image

Return values

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

◆ tivxDLDrawBoxNode()

VX_API_ENTRY vx_node VX_API_CALL tivxDLDrawBoxNode	(	vx_graph	graph,
		vx_user_data_object	config,
		vx_image	input_image,
		vx_tensor	input_tensor,
		vx_image	output_image
	)

DL Draw Box Node.

Takes the DL output and draws boxes around an object on input image

Parameters

[in]	graph	reference to the graph
[in]	config	reference to tivxDLPreProcPrams
[in]	input_image	Input image
[in]	input_tensor	Input tensor
[out]	output_image	Output image

Return values

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

◆ tivxDLColorConvertNode()

VX_API_ENTRY vx_node VX_API_CALL tivxDLColorConvertNode	(	vx_graph	graph,
		vx_image	input_image,
		vx_image	output_image
	)

DL Color Convert Node.

Performs limited color conversions required for DL pipeline which is DMA optimized

Parameters

[in]	graph	reference to the graph
[in]	input_image	Input image
[out]	output_image	Output image

Return values

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

Introduction

Data Structures

Functions

Macros

Macro Definition Documentation

◆ TIVX_MODULE_NAME_IMG_PROC

◆ TIVX_PIXEL_VIZ_MAX_TENSOR

Function Documentation

◆ tivxImgProcLoadKernels()

◆ tivxImgProcUnLoadKernels()

◆ tivxAddKernelPixelViz()

◆ tivxAddKernelImgMosaic()

◆ tivxImgMosaicParamsSetDefaults()

◆ tivxImgPreProcNode()

◆ tivxOCPreProcNode()

◆ tivxOCPostProcNode()

◆ tivxODPostProcNode()

◆ tivxDofPlaneSepNode()

◆ tivxPixelVizNode()

◆ tivxVisualLocalizationNode()

◆ tivxPoseVizNode()

◆ tivxImgMosaicNode()

◆ tivxDrawKeypointDetectionsNode()

◆ tivxDrawBoxDetectionsNode()

◆ tivxDLBEVPostProcNode()

◆ tivxImgHistNode()

◆ tivxSFMNode()

◆ tivxDLPreProcNode()

◆ tivxDLColorBlendNode()

◆ tivxDLDrawBoxNode()

◆ tivxDLColorConvertNode()