Histogram of Oriented Gradients

Extracts Histogram of Oriented Gradients features from input image. More...

Functions
VPIStatus	vpiCreateExtractHOGFeatures (uint64_t backends, int32_t width, int32_t height, int32_t features, int32_t cellSize, int32_t numOrientations, int32_t *outNumFeatures, VPIPayload *payload)
	Create a payload for the non-batch version of HOG algorithm. More...
VPIStatus	vpiCreateExtractHOGFeaturesBatch (uint64_t backends, int32_t maxBatchWidth, int32_t maxBatchHeight, int32_t imgWidth, int32_t imgHeight, int32_t features, int32_t cellSize, int32_t numOrientations, int32_t *outNumFeatures, VPIPayload *payload)
	Create a payload for the batch version of HOG algorithm. More...
VPIStatus	vpiSubmitExtractHOGFeatures (VPIStream stream, uint64_t backend, VPIPayload payload, VPIImage input, VPIImage *outFeatures, int32_t numFeatures)
	Submits a HOG operation to the stream. More...

HOG flags
Defines what features will be returned.
#define	VPI_HOG_CONTRAST_SENSITIVE   0x01
	Return contrast sensitive features.
#define	VPI_HOG_CONTRAST_INSENSITIVE   0x02
	Return contrast insensitive features.
#define	VPI_HOG_TEXTURE   0x04
	Return texture-related features.
#define	VPI_HOG_ALL_FEATURES   (VPI_HOG_CONTRAST_SENSITIVE | VPI_HOG_CONTRAST_INSENSITIVE | VPI_HOG_TEXTURE)
	Helper flag to return all features.

Detailed Description

Extracts Histogram of Oriented Gradients features from input image.

Function Documentation

vpiCreateExtractHOGFeatures()

VPIStatus vpiCreateExtractHOGFeatures	(	uint64_t	backends,
		int32_t	width,
		int32_t	height,
		int32_t	features,
		int32_t	cellSize,
		int32_t	numOrientations,
		int32_t *	outNumFeatures,
		VPIPayload *	payload
	)

#include </opt/nvidia/vpi2/include/vpi/experimental/HOG.h>

Create a payload for the non-batch version of HOG algorithm.

Parameters

[in]	backends	VPI backends that are eligible to execute the algorithm. Valid values: VPI_BACKEND_CPU VPI_BACKEND_CUDA
[in]	width,height	Dimensions of the input image to be used. Must be >= 0.
[in]	features	Flags to specify what features will be returned. Must be a bitwise combination of one or more of the following flags: VPI_HOG_CONTRAST_SENSITIVE VPI_HOG_CONTRAST_INSENSITIVE VPI_HOG_TEXTURE VPI_HOG_ALL_FEATURES (it's the combination of all flags above)
[in]	cellSize	Cell size, typically 8 or 16 for 8x8 and 16x16 cells respectively. Must be >=2 and <= 32 and power of two.
[in]	numOrientations	Number of orientations used. This is typically 18. Must be between 4 and 18.
[out]	outNumFeatures	Receives the number of features that will be returned. Pass NULL if not needed.
[out]	payload	Pointer to the payload variable that receives the created handle.

Return values

VPI_ERROR_INVALID_ARGUMENT	payload handle is NULL.
VPI_ERROR_INVALID_ARGUMENT	width or height outside valid range.
VPI_ERROR_NOT_IMPLEMENTED	HOG is not implemented for given backend.
VPI_ERROR_INVALID_CONTEXT	Current context is destroyed.
VPI_ERROR_OUT_OF_MEMORY	Cannot allocate required resources.
VPI_SUCCESS	Operation executed successfully.

vpiCreateExtractHOGFeaturesBatch()

VPIStatus vpiCreateExtractHOGFeaturesBatch	(	uint64_t	backends,
		int32_t	maxBatchWidth,
		int32_t	maxBatchHeight,
		int32_t	imgWidth,
		int32_t	imgHeight,
		int32_t	features,
		int32_t	cellSize,
		int32_t	numOrientations,
		int32_t *	outNumFeatures,
		VPIPayload *	payload
	)

#include </opt/nvidia/vpi2/include/vpi/experimental/HOG.h>

Create a payload for the batch version of HOG algorithm.

The input images are assumed to be laid out in memory as a 2D matrix of images. All images must have the same dimensions.

Parameters

[in]	backends	VPI backends that are eligible to execute the algorithm. Valid values: VPI_BACKEND_CPU VPI_BACKEND_CUDA
[in]	maxBatchWidth,maxBatchHeight	Maximum number of images horizontally and vertically, respectively. Must be >= 0.
[in]	imgWidth,imgHeight	Dimensions each of the batch images. Must be >= 0. imgWidth and imgHeight must be multiple of cell width and height respectively.
[in]	features	Flags to specify which features will be returned. Must be a bitwise combination of one or more of the following flags: VPI_HOG_CONTRAST_SENSITIVE VPI_HOG_CONTRAST_INSENSITIVE VPI_HOG_TEXTURE VPI_HOG_ALL_FEATURES (it's the combination of all flags above)
[in]	cellSize	Cell size, typically 8 or 16 for 8x8 and 16x16 cells respectively. Must be >=2 and <= 32 and power of two.
[in]	numOrientations	Number of orientations used. This is typically 18. Must be between 4 and 18.
[out]	outNumFeatures	Receives the number of features that will be returned. Pass NULL if not needed.
[out]	payload	Pointer to a payload handle that will receive the allocated payload.

Return values

VPI_ERROR_INVALID_ARGUMENT	payload handle is NULL.
VPI_ERROR_INVALID_ARGUMENT	maxBatchWidth or maxBatchHeight outside valid range.
VPI_ERROR_INVALID_ARGUMENT	imgWidth or imgHeight outside valid range.
VPI_ERROR_INVALID_ARGUMENT	Invalid features.
VPI_ERROR_INVALID_ARGUMENT	cellSize or numOrientations outside valid range.
VPI_ERROR_INVALID_ARGUMENT	imgWidth must be a multiple of cell width.
VPI_ERROR_INVALID_ARGUMENT	imgHeight must be a multiple of cell height.
VPI_ERROR_INVALID_ARGUMENT	backends refers to an invalid backend.
VPI_ERROR_NOT_IMPLEMENTED	HOG is not implemented for given backend.
VPI_ERROR_INVALID_CONTEXT	Current context is destroyed.
VPI_ERROR_OUT_OF_MEMORY	Cannot allocate required resources.
VPI_SUCCESS	Operation executed successfully.

vpiSubmitExtractHOGFeatures()

VPIStatus vpiSubmitExtractHOGFeatures	(	VPIStream	stream,
		uint64_t	backend,
		VPIPayload	payload,
		VPIImage	input,
		VPIImage *	outFeatures,
		int32_t	numFeatures
	)

#include </opt/nvidia/vpi2/include/vpi/experimental/HOG.h>

Submits a HOG operation to the stream.

It handles both batch and non-batch payloads.

Parameters

[in]	stream	The stream where the operation will be queued in. Must not be NULL. Stream must have enabled the backends that will execute the algorithm.
[in]	backend	Backend that will execute the algorithm. Must be the backend specified during payload creation or 0 as a shorthand to use this backend.
[in]	payload	Payload to be submitted along the other parameters.
[in]	input	If using a non batch payload, this is the input image to be processed. Must not be NULL. Its dimensions must match what was passed to vpiCreateExtractHOGFeatures. If using a batch payload, the image dimensions must be a multiple of (imgWidth,imgHeight), and it must be at most (imgWidth*maxBatchWidth, imgHeight*maxBatchHeight). Image must have enabled the backends that will execute the algorithm.
[out]	outFeatures	Pointer to an array of images that will receive the features. Must not be NULL. The dimensions of each image must be (input.width / cellSize, input.height / cellSize). In case of batch processing, the output features position in the 2D matrix be the same position of the corresponding input image. All images must have same format. All image must have enabled the backends that will execute the algorithm.
[in]	numFeatures	Number of images in the output array. Must be between 1 and 32.

Return values

VPI_ERROR_INVALID_ARGUMENT	stream is NULL.
VPI_ERROR_INVALID_ARGUMENT	input or outFeatures are NULL.
VPI_ERROR_INVALID_ARGUMENT	payload is not generated using vpiCreateExtractHOGFeatures.
VPI_ERROR_INVALID_ARGUMENT	numFeatures outside valid range.
VPI_ERROR_INVALID_ARGUMENT	Batch input width must be a multiple of configured input width.
VPI_ERROR_INVALID_ARGUMENT	Batch input height must be a multiple of configured input height.
VPI_ERROR_INVALID_ARGUMENT	outFeatures must all have the same format.
VPI_ERROR_INVALID_PAYLOAD_TYPE	payload is invalid.
VPI_ERROR_INVALID_OPERATION	The needed backends aren't enabled in stream, input or outFeatures.
VPI_SUCCESS	Operation executed successfully.