Detailed Description

Runs the stereo pipeline and computes disparity map on PVA.

Runs the stereo pipeline's individual steps and computes disparity map on PVA.

Note: SW Release Applicability: These APIs are available in both NVIDIA DriveWorks and NVIDIA DRIVE Software releases.

Data Structures
struct	dwStereoPVAParams
	Configuration parameters for a Stereo algorithm. More...

Typedefs
typedef struct dwStereoPVAObject *	dwStereoPVAHandle_t
	A pointer to the handle representing a stereo algorithm on PVA. More...

Enumerations
enum	dwStereoPVAMode { DW_STEREO_PVA_MODE_MEDIUM_QUALITY = 0, DW_STEREO_PVA_MODE_LOW_QUALITY = 1, DW_STEREO_PVA_MODE_HIGH_QUALITY = 2 }
	Stereo computation modes. More...

enum	dwStereoStage { DW_STEREO_STAGE_CPU_SYNC_CONVERT_DATA = 0, DW_STEREO_STAGE_VIC_ASYNC_CONVERT_DATA = 1, DW_STEREO_STAGE_PVA_ASYNC_PREPROCESS = 2, DW_STEREO_STAGE_NVENC_ASYNC_PROCESS = 3, DW_STEREO_STAGE_PVA_ASYNC_POSTPROCESS = 4, DW_STEREO_STAGE_GPU_ASYNC_POSTPROCESS = 5, DW_STEREO_STAGE_CPU_SYNC_PROCESS = 6 }
	Defines stages of the pipleine. More...

Functions
DW_API_PUBLIC dwStatus	dwStereoPVA_bindInput (dwImageHandle_t left, dwImageHandle_t right, dwStereoPVAHandle_t obj)
	Binds the input images to stereo module. More...

DW_API_PUBLIC dwStatus	dwStereoPVA_bindOutput (dwImageHandle_t disparity, dwImageHandle_t confidence, dwStereoPVAHandle_t obj)
	Binds the output images to stereo module. More...

DW_API_PUBLIC dwStatus	dwStereoPVA_computeDisparity (dwImageHandle_t disparityMap, dwImageHandle_t confidenceMap, const dwImageHandle_t leftImage, const dwImageHandle_t rightImage, dwStereoPVAHandle_t obj)
	Executes the full pipeline and computes disparity map. More...

DW_API_PUBLIC dwStatus	dwStereoPVA_getConfidenceImageProperties (dwImageProperties *props, dwStereoPVAHandle_t obj)
	Get output confidence image properties. More...

DW_API_PUBLIC dwStatus	dwStereoPVA_getCUDAStream (cudaStream_t *stream, dwStereoPVAHandle_t obj)
	Gets CUDA stream used by the stereo algorithm. More...

DW_API_PUBLIC dwStatus	dwStereoPVA_getDisparityImageProperties (dwImageProperties *props, dwStereoPVAHandle_t obj)
	Get output disparity image properties. More...

DW_API_PUBLIC dwStatus	dwStereoPVA_initialize (dwStereoPVAHandle_t obj, uint32_t width, uint32_t height, const dwStereoPVAParams stereoParams, dwContextHandle_t ctx)
	Initializes the stereo algorithm with the parameters. More...

DW_API_PUBLIC dwStatus	dwStereoPVA_initParams (dwStereoPVAParams *stereoParams)
	Initializes the StereoPVA parameters. More...

DW_API_PUBLIC dwStatus	dwStereoPVA_process (dwStereoStage stage, dwStereoPVAHandle_t obj)
	Executes a given stage of the pipeline. More...

DW_API_PUBLIC dwStatus	dwStereoPVA_release (dwStereoPVAHandle_t obj)
	Releases the stereo algorithm. More...

DW_API_PUBLIC dwStatus	dwStereoPVA_reset (dwStereoPVAHandle_t obj)
	Resets the Stereo module. More...

DW_API_PUBLIC dwStatus	dwStereoPVA_setCUDAStream (cudaStream_t stream, dwStereoPVAHandle_t obj)
	Sets CUDA stream used by the stereo algorithm. More...

Data Structure Documentation

◆ dwStereoPVAParams

struct dwStereoPVAParams

Data Fields
dwStereoPVAMode	estimationMode	Estimation mode that determines the speed and quality of stereo algorithm.
dwProcessorType	processorNVENC	Specifies the processor type for the stages of the pipeline which run on NVENC. Supported options are DW_PROCESSOR_TYPE_NVENC_0, DW_PROCESSOR_TYPE_NVENC_1
dwProcessorType	processorPVA	Specifies the processor type for the stages of the pipeline which run on PVA. Supported options are DW_PROCESSOR_TYPE_PVA_0, DW_PROCESSOR_TYPE_PVA_1

Typedef Documentation

◆ dwStereoPVAHandle_t

typedef struct dwStereoPVAObject* dwStereoPVAHandle_t

A pointer to the handle representing a stereo algorithm on PVA.

This object allows the computation of a disparity map given two rectified stereo images

Definition at line 65 of file StereoPVA.h.

Enumeration Type Documentation

◆ dwStereoPVAMode

enum dwStereoPVAMode

Stereo computation modes.

Enumerator
DW_STEREO_PVA_MODE_MEDIUM_QUALITY	High performance with medium quality.
DW_STEREO_PVA_MODE_LOW_QUALITY	Ultra high performance with lowest quality.
DW_STEREO_PVA_MODE_HIGH_QUALITY	High quality with low performance.

Definition at line 70 of file StereoPVA.h.

◆ dwStereoStage

enum dwStereoStage

Defines stages of the pipleine.

Enumerator
DW_STEREO_STAGE_CPU_SYNC_CONVERT_DATA	Convert format of the image on CPU. This stage is only executed if input images are NvMedia. It's NOOP otherwise.
DW_STEREO_STAGE_VIC_ASYNC_CONVERT_DATA	Prepares data by converting pitch linear NvMedia image to block linear.
DW_STEREO_STAGE_PVA_ASYNC_PREPROCESS	Generates hints for stereo disparity computation by executing the algorithm on PVA on the downscaled image.
DW_STEREO_STAGE_NVENC_ASYNC_PROCESS	Computes disparity on NVENC.
DW_STEREO_STAGE_PVA_ASYNC_POSTPROCESS	Refines the output of DW_STEREO_STAGE_NVENC_ASYNC_PROCESS based on the confidence map received from the DW_STEREO_STAGE_PVA_ASYNC_PREPROCESS.
DW_STEREO_STAGE_GPU_ASYNC_POSTPROCESS	Schedules extracting disparity information and streaming of data from PVA back to GPU.
DW_STEREO_STAGE_CPU_SYNC_PROCESS	A blocking stage to ensure that the computation is finalized.

Definition at line 58 of file StereoPVA_processpipeline.h.

Function Documentation

◆ dwStereoPVA_bindInput()

DW_API_PUBLIC dwStatus dwStereoPVA_bindInput	(	dwImageHandle_t	left,
		dwImageHandle_t	right,
		dwStereoPVAHandle_t	obj
	)

Binds the input images to stereo module.

Parameters

[in]	left	Left image of format DW_IMAGE_FORMAT_YUV420_UINT8_SEMIPLANAR
[in]	right	Right image of format DW_IMAGE_FORMAT_YUV420_UINT8_SEMIPLANAR
[in]	obj	Specifies the StereoPVA handle.

Returns: DW_SUCCESS, DW_INVALID_HANDLE, DW_BAD_CAST

Note: If an image is provided in CUDA, it'll be streamed to NvMedia internally.; If an image is provided in NvMedia, the first stage will be executed

◆ dwStereoPVA_bindOutput()

DW_API_PUBLIC dwStatus dwStereoPVA_bindOutput	(	dwImageHandle_t	disparity,
		dwImageHandle_t	confidence,
		dwStereoPVAHandle_t	obj
	)

Binds the output images to stereo module.

Parameters

[in]	disparity	Disparity image of format DW_IMAGE_FORMAT_R_INT16
[in]	confidence	Confidence image of format DW_IMAGE_FORMAT_R_UINT16
[in]	obj	Specifies the StereoPVA handle.

Returns: DW_SUCCESS, DW_INVALID_HANDLE, DW_BAD_CAST

Note: If an image is provided in CUDA, it'll be streamed to NvMedia internally.; If an image is provided in NvMedia, the first stage will be executed

◆ dwStereoPVA_computeDisparity()

DW_API_PUBLIC dwStatus dwStereoPVA_computeDisparity	(	dwImageHandle_t	disparityMap,
		dwImageHandle_t	confidenceMap,
		const dwImageHandle_t	leftImage,
		const dwImageHandle_t	rightImage,
		dwStereoPVAHandle_t	obj
	)

Executes the full pipeline and computes disparity map.

The pipeline runs on the following engines:

VIC (scaling & converting data)
PVA (hint generation)
NVENC (stereo disparity computation)
PVA (refining results)
GPU (copying memory back to GPU)

Parameters

[out]	disparityMap	Disparity map
[out]	confidenceMap	Confidence map
[in]	leftImage	Left image. Expected YUV
[in]	rightImage	The left 8 bit rectified image pyramid.
[in]	obj	The stereo algorithm handle.

Returns: DW_NVMEDIA_ERROR - if the underlying stereo algorithm had an internal error.
DW_INTERNAL_ERROR - if the underlying stereo algorithm had an internal error.
DW_INVALID_HANDLE - if given handle is not valid, i.e. null or of wrong type .
DW_SUCCESS

◆ dwStereoPVA_getConfidenceImageProperties()

DW_API_PUBLIC dwStatus dwStereoPVA_getConfidenceImageProperties	(	dwImageProperties *	props,
		dwStereoPVAHandle_t	obj
	)

Get output confidence image properties.

Parameters

[out]	props
[in]	obj

Returns: DW_INVALID_HANDLE if the given context handle is invalid,i.e. null or of wrong type
or DW_SUCCESS otherwise.

◆ dwStereoPVA_getCUDAStream()

DW_API_PUBLIC dwStatus dwStereoPVA_getCUDAStream	(	cudaStream_t *	stream,
		dwStereoPVAHandle_t	obj
	)

Gets CUDA stream used by the stereo algorithm.

Parameters

[out]	stream	The CUDA stream currently used by the stereo algorithm.
[in]	obj	The stereo algorithm handle.

Returns: DW_INVALID_HANDLE if the given context handle is invalid, i.e. null or of wrong type
or DW_SUCCESS otherwise.

◆ dwStereoPVA_getDisparityImageProperties()

DW_API_PUBLIC dwStatus dwStereoPVA_getDisparityImageProperties	(	dwImageProperties *	props,
		dwStereoPVAHandle_t	obj
	)

Get output disparity image properties.

Parameters

[out]	props
[in]	obj

Returns: DW_INVALID_HANDLE if the given context handle is invalid,i.e. null or of wrong type
or DW_SUCCESS otherwise.

◆ dwStereoPVA_initialize()

DW_API_PUBLIC dwStatus dwStereoPVA_initialize	(	dwStereoPVAHandle_t *	obj,
		uint32_t	width,
		uint32_t	height,
		const dwStereoPVAParams *	stereoParams,
		dwContextHandle_t	ctx
	)

Initializes the stereo algorithm with the parameters.

Parameters

[out]	obj	A pointer to the stereo algorithm on PVA.
[in]	width	The width of one input image.
[in]	height	The height of one input image.
[in]	stereoParams	A pointer to the configuration of the stereo algorithm.
[in]	ctx	the handle to DW context.

Returns: DW_CUDA_ERROR - if the underlying stereo algorithm had a CUDA error.
DW_INVALID_HANDLE - if given handle is not valid, i.e. null or of wrong type .
DW_SUCCESS

◆ dwStereoPVA_initParams()

DW_API_PUBLIC dwStatus dwStereoPVA_initParams ( dwStereoPVAParams * stereoParams )

Initializes the StereoPVA parameters.

Parameters

[out] stereoParams Parameters to be initialised with default values.

Returns: DW_INVALID_HANDLE - if given handle is not valid, i.e. null or of wrong type .
DW_SUCCESS

◆ dwStereoPVA_process()

DW_API_PUBLIC dwStatus dwStereoPVA_process	(	dwStereoStage	stage,
		dwStereoPVAHandle_t	obj
	)

Executes a given stage of the pipeline.

Parameters

[in]	stage	Stage to execute.
[in]	obj	Specifies the StereoPVA handle.

Note: The stages must be executed in the order of the enumeration.

Returns: DW_SUCCESS, DW_INVALID_HANDLE, DW_BAD_CAST, DW_INVALID_ARGUMENT

◆ dwStereoPVA_release()

DW_API_PUBLIC dwStatus dwStereoPVA_release ( dwStereoPVAHandle_t obj )

Releases the stereo algorithm.

This method releases all resources associated with a stereo algorithm.

Note: This method renders the handle unusable.

Parameters

[in] obj The object handle to be released.

Returns: DW_SUCCESS
DW_INVALID_HANDLE - If the given context handle is invalid,i.e. null or of wrong type
DW_BAD_CAST

◆ dwStereoPVA_reset()

DW_API_PUBLIC dwStatus dwStereoPVA_reset ( dwStereoPVAHandle_t obj )

Resets the Stereo module.

Parameters

[in] obj Specifies the stereo handle to reset.

Returns: DW_SUCCESS
DW_INVALID_HANDLE - If the given context handle is invalid,i.e. null or of wrong type
DW_BAD_CAST

◆ dwStereoPVA_setCUDAStream()

DW_API_PUBLIC dwStatus dwStereoPVA_setCUDAStream	(	cudaStream_t	stream,
		dwStereoPVAHandle_t	obj
	)

Sets CUDA stream used by the stereo algorithm.

Parameters

[in]	stream	The CUDA stream.
[in]	obj	The stereo algorithm handle.

Returns: DW_INVALID_HANDLE if the given context handle is invalid,i.e. null or of wrong type
or DW_SUCCESS otherwise.

Detailed Description

Data Structures

Typedefs

Enumerations

Functions

Data Structure Documentation

◆ dwStereoPVAParams

Typedef Documentation

◆ dwStereoPVAHandle_t

Enumeration Type Documentation

◆ dwStereoPVAMode

◆ dwStereoStage

Function Documentation

◆ dwStereoPVA_bindInput()

◆ dwStereoPVA_bindOutput()

◆ dwStereoPVA_computeDisparity()

◆ dwStereoPVA_getConfidenceImageProperties()

◆ dwStereoPVA_getCUDAStream()

◆ dwStereoPVA_getDisparityImageProperties()

◆ dwStereoPVA_initialize()

◆ dwStereoPVA_initParams()

◆ dwStereoPVA_process()

◆ dwStereoPVA_release()

◆ dwStereoPVA_reset()

◆ dwStereoPVA_setCUDAStream()