Detailed Description

Class of the nvinferserver element implementation.

Definition at line 142 of file gstnvinferserver_impl.h.

Collaboration diagram for gstnvinferserver::GstNvInferServerImpl:

Public Member Functions
	GstNvInferServerImpl (GstNvInferServer *infer)
	Constructor, registers the handle of the parent GStreamer element. More...

	~GstNvInferServerImpl ()
	Constructor, do nothing. More...

void	setRawoutputCb (RawOutputCallback cb)
	Saves the callback function pointer for the raw tensor output. More...

NvDsInferStatus	start ()
	Reads the configuration file and sets up processing context. More...

NvDsInferStatus	stop ()
	Deletes the inference context. More...

NvDsInferStatus	processBatchMeta (NvDsBatchMeta batchMeta, NvBufSurface inSurf, uint64_t seqId, void *gstBuf)
	Submits the input batch for inference. More...

NvDsInferStatus	sync ()
	Waits for the output thread to finish processing queued operations. More...

NvDsInferStatus	queueOperation (FuncItem func)
	Queues the inference done operation for the request to the output thread. More...

bool	addTrackingSource (uint32_t sourceId)
	Add a new source to the object history structure. More...

void	eraseTrackingSource (uint32_t sourceId)
	Removes a source from the object history structure. More...

void	resetIntervalCounter ()
	Resets the inference interval used in frame process mode to 0. More...

uint32_t	uniqueId () const

const std::string &	classifierType () const

uint32_t	maxBatchSize () const

nvtxDomainHandle_t	nvtxDomain ()

bool	isAsyncMode () const

bool	canSupportGpu (int gpuId) const

NvDsInferStatus	lastError () const

const ic::PluginControl &	config () const

void	updateInterval (guint interval)

Data Fields
GstNvInferServerProperties	m_GstProperties

Constructor & Destructor Documentation

◆ GstNvInferServerImpl()

gstnvinferserver::GstNvInferServerImpl::GstNvInferServerImpl ( GstNvInferServer * infer )

Constructor, registers the handle of the parent GStreamer element.

Parameters

[in] infer Pointer to the nvinferserver GStreamer element.

◆ ~GstNvInferServerImpl()

gstnvinferserver::GstNvInferServerImpl::~GstNvInferServerImpl ( )

Constructor, do nothing.

Member Function Documentation

◆ addTrackingSource()

bool gstnvinferserver::GstNvInferServerImpl::addTrackingSource ( uint32_t sourceId )

Add a new source to the object history structure.

Whenever a new source is added to the pipeline, corresponding source ID is captured in the GStreamer event on the sink pad and the object history for this source is initialized.

Parameters

[in] id Source ID.

◆ canSupportGpu()

bool gstnvinferserver::GstNvInferServerImpl::canSupportGpu ( int gpuId ) const

◆ classifierType()

const std::string& gstnvinferserver::GstNvInferServerImpl::classifierType ( ) const

inline

Definition at line 349 of file gstnvinferserver_impl.h.

References gstnvinferserver::GstNvInferServerProperties::classifierType, and m_GstProperties.

◆ config()

const ic::PluginControl& gstnvinferserver::GstNvInferServerImpl::config ( ) const

inline

Definition at line 357 of file gstnvinferserver_impl.h.

◆ eraseTrackingSource()

void gstnvinferserver::GstNvInferServerImpl::eraseTrackingSource ( uint32_t sourceId )

Removes a source from the object history structure.

Whenever a source is removed from the pipeline, corresponding source ID is captured in the GStreamer event on the sink pad and the object history for this source is deleted.

Parameters

[in] id Source ID.

◆ isAsyncMode()

bool gstnvinferserver::GstNvInferServerImpl::isAsyncMode ( ) const

◆ lastError()

NvDsInferStatus gstnvinferserver::GstNvInferServerImpl::lastError ( ) const

◆ maxBatchSize()

uint32_t gstnvinferserver::GstNvInferServerImpl::maxBatchSize ( ) const

◆ nvtxDomain()

nvtxDomainHandle_t gstnvinferserver::GstNvInferServerImpl::nvtxDomain ( )

inline

Definition at line 351 of file gstnvinferserver_impl.h.

◆ processBatchMeta()

NvDsInferStatus gstnvinferserver::GstNvInferServerImpl::processBatchMeta	(	NvDsBatchMeta *	batchMeta,
		NvBufSurface *	inSurf,
		uint64_t	seqId,
		void *	gstBuf
	)

Submits the input batch for inference.

This function submits the input batch buffer for inferencing as per the configured processing mode: full frame inference, inference on detected objects or inference on attached input tensors.

Parameters

[in,out]	batchMeta	NvDsBatchMeta associated with the input buffer.
[in]	inSurf	Input batch buffer.
[in]	seqId	The sequence number of the input batch.
[in]	gstBuf	Pointer to the input GStreamer buffer.

◆ queueOperation()

NvDsInferStatus gstnvinferserver::GstNvInferServerImpl::queueOperation ( FuncItem func )

Queues the inference done operation for the request to the output thread.

◆ resetIntervalCounter()

void gstnvinferserver::GstNvInferServerImpl::resetIntervalCounter ( )

Resets the inference interval used in frame process mode to 0.

◆ setRawoutputCb()

void gstnvinferserver::GstNvInferServerImpl::setRawoutputCb ( RawOutputCallback cb )

inline

Saves the callback function pointer for the raw tensor output.

Parameters

[in] cb Pointer to the callback function.

Definition at line 264 of file gstnvinferserver_impl.h.

◆ start()

NvDsInferStatus gstnvinferserver::GstNvInferServerImpl::start ( )

Reads the configuration file and sets up processing context.

This function reads the configuration file and validates the user provided configuration. Configuration file settings are overridden with those set by element properties. It then creates the inference context, initializes it and starts the output thread. The inference context is either InferGrpcContext (Triton Inference Server in gRPC mode) or InferTrtISContext (Triton Inference server C-API mode) depending on the configuration setting. Object history is initialized for source 0.

◆ stop()

NvDsInferStatus gstnvinferserver::GstNvInferServerImpl::stop ( )

Deletes the inference context.

This function waits for the output thread to finish and then de-initializes the inference context. The object history is cleared.

◆ sync()

NvDsInferStatus gstnvinferserver::GstNvInferServerImpl::sync ( )

Waits for the output thread to finish processing queued operations.

◆ uniqueId()

uint32_t gstnvinferserver::GstNvInferServerImpl::uniqueId ( ) const

inline

Definition at line 348 of file gstnvinferserver_impl.h.

References m_GstProperties, and gstnvinferserver::GstNvInferServerProperties::uniqueId.

◆ updateInterval()

void gstnvinferserver::GstNvInferServerImpl::updateInterval ( guint interval )

inline

Definition at line 358 of file gstnvinferserver_impl.h.

Field Documentation

◆ m_GstProperties

GstNvInferServerProperties gstnvinferserver::GstNvInferServerImpl::m_GstProperties

Definition at line 548 of file gstnvinferserver_impl.h.

Referenced by classifierType(), and uniqueId().

The documentation for this class was generated from the following file:

gstnvinferserver_impl.h

NVIDIA DeepStream SDK API Reference

6.4 Release

Detailed Description

Public Member Functions

Data Fields

Constructor & Destructor Documentation

◆ GstNvInferServerImpl()

◆ ~GstNvInferServerImpl()

Member Function Documentation

◆ addTrackingSource()

◆ canSupportGpu()

◆ classifierType()

◆ config()

◆ eraseTrackingSource()

◆ isAsyncMode()

◆ lastError()

◆ maxBatchSize()

◆ nvtxDomain()

◆ processBatchMeta()

◆ queueOperation()

◆ resetIntervalCounter()

◆ setRawoutputCb()

◆ start()

◆ stop()

◆ sync()

◆ uniqueId()

◆ updateInterval()

Field Documentation

◆ m_GstProperties