NVIDIA DeepStream SDK API Reference

9.2 Release
IPoseEstimator Class Referenceabstract

Detailed Description

Defines the interface for per-target pose estimation.

Definition at line 35 of file sources/libs/nvmultiobjecttracker/include/modules/PoseEstimator/IPoseEstimator.hpp.

Inheritance diagram for IPoseEstimator:

Public Member Functions

virtual ~IPoseEstimator ()
 Release resources owned by the IPoseEstimator instance. More...
 
virtual void initializeRuntime (const SourceFrameInfo &frameInfo, const uint maxTargetsPerStream, const std::shared_ptr< NvTrackedObjectManager > &targetManager)=0
 Initialize variables and allocate memories. More...
 
virtual void initializeModel ()=0
 Create TensorRT inference engine; if engine does not exist, serialize from model file. More...
 
virtual void queryParams (NvMOTQuery &paramsQuery)=0
 Populate the capabilities reported to the tracker. More...
 
virtual void doInference (const std::map< NvMOTStreamId, NvMOTFrame * > &inputMeta, std::map< NvMOTStreamId, std::deque< DetectedObjInfo >> &detectedObjects, const bool isForReAssoc)=0
 Run pose inference on cropped targets from the input frames. More...
 
virtual void update ()=0
 Update the pose estimator after data association. More...
 
virtual void removeStream (const NvMOTStreamId videoStreamId)=0
 Remove a stream from the pose estimator. More...
 
virtual const POSEESTIMATOR_TYPES getPoseEstimatorType ()=0
 Get the configured pose-estimator type. More...
 
virtual ~IPoseEstimator ()
 Release resources owned by the IPoseEstimator instance. More...
 
virtual void initializeRuntime (const SourceFrameInfo &frameInfo, const uint maxTargetsPerStream, const std::shared_ptr< NvTrackedObjectManager > &targetManager)=0
 Initialize variables and allocate memories. More...
 
virtual void initializeModel ()=0
 Create TensorRT inference engine; if engine does not exist, serialize from model file. More...
 
virtual void queryParams (NvMOTQuery &paramsQuery)=0
 Populate the capabilities reported to the tracker. More...
 
virtual void doInference (const std::map< NvMOTStreamId, NvMOTFrame * > &inputMeta, std::map< NvMOTStreamId, std::deque< DetectedObjInfo >> &detectedObjects, const bool isForReAssoc)=0
 Run pose inference on cropped targets from the input frames. More...
 
virtual void update ()=0
 Update the pose estimator after data association. More...
 
virtual void removeStream (const NvMOTStreamId videoStreamId)=0
 Remove a stream from the pose estimator. More...
 
virtual const POSEESTIMATOR_TYPES getPoseEstimatorType ()=0
 Get the configured pose-estimator type. More...
 

Constructor & Destructor Documentation

◆ ~IPoseEstimator() [1/2]

virtual IPoseEstimator::~IPoseEstimator ( )
inlinevirtual

Release resources owned by the IPoseEstimator instance.

Definition at line 39 of file sources/libs/nvmultiobjecttracker/include/modules/PoseEstimator/IPoseEstimator.hpp.

◆ ~IPoseEstimator() [2/2]

virtual IPoseEstimator::~IPoseEstimator ( )
inlinevirtual

Release resources owned by the IPoseEstimator instance.

Definition at line 39 of file 9.2/sources/libs/nvmultiobjecttracker/include/modules/PoseEstimator/IPoseEstimator.hpp.

Member Function Documentation

◆ doInference() [1/2]

virtual void IPoseEstimator::doInference ( const std::map< NvMOTStreamId, NvMOTFrame * > &  inputMeta,
std::map< NvMOTStreamId, std::deque< DetectedObjInfo >> &  detectedObjects,
const bool  isForReAssoc 
)
pure virtual

Run pose inference on cropped targets from the input frames.

Parameters
[in]inputMetaA batch of input frames.
[in,out]detectedObjectsNewly detected objects in each stream.
[in]isForReAssocWhether this inference is for reassociation.

Implemented in PoseInference, and PoseInference.

◆ doInference() [2/2]

virtual void IPoseEstimator::doInference ( const std::map< NvMOTStreamId, NvMOTFrame * > &  inputMeta,
std::map< NvMOTStreamId, std::deque< DetectedObjInfo >> &  detectedObjects,
const bool  isForReAssoc 
)
pure virtual

Run pose inference on cropped targets from the input frames.

Parameters
[in]inputMetaA batch of input frames.
[in,out]detectedObjectsNewly detected objects in each stream.
[in]isForReAssocWhether this inference is for reassociation.

Implemented in PoseInference, and PoseInference.

◆ getPoseEstimatorType() [1/2]

virtual const POSEESTIMATOR_TYPES IPoseEstimator::getPoseEstimatorType ( )
pure virtual

Get the configured pose-estimator type.

Implemented in PoseInference, and PoseInference.

◆ getPoseEstimatorType() [2/2]

virtual const POSEESTIMATOR_TYPES IPoseEstimator::getPoseEstimatorType ( )
pure virtual

Get the configured pose-estimator type.

Implemented in PoseInference, and PoseInference.

◆ initializeModel() [1/2]

virtual void IPoseEstimator::initializeModel ( )
pure virtual

Create TensorRT inference engine; if engine does not exist, serialize from model file.

Implemented in PoseInference, and PoseInference.

◆ initializeModel() [2/2]

virtual void IPoseEstimator::initializeModel ( )
pure virtual

Create TensorRT inference engine; if engine does not exist, serialize from model file.

Implemented in PoseInference, and PoseInference.

◆ initializeRuntime() [1/2]

virtual void IPoseEstimator::initializeRuntime ( const SourceFrameInfo &  frameInfo,
const uint  maxTargetsPerStream,
const std::shared_ptr< NvTrackedObjectManager > &  targetManager 
)
pure virtual

Initialize variables and allocate memories.

Parameters
[in]frameInfoMetadata about the source image, such as size and color.
[in]maxTargetsPerStreamMaximum number of targets to track in each stream.
[in]targetManagerTarget manager of the tracked objects.

Implemented in PoseInference, and PoseInference.

◆ initializeRuntime() [2/2]

virtual void IPoseEstimator::initializeRuntime ( const SourceFrameInfo &  frameInfo,
const uint  maxTargetsPerStream,
const std::shared_ptr< NvTrackedObjectManager > &  targetManager 
)
pure virtual

Initialize variables and allocate memories.

Parameters
[in]frameInfoMetadata about the source image, such as size and color.
[in]maxTargetsPerStreamMaximum number of targets to track in each stream.
[in]targetManagerTarget manager of the tracked objects.

Implemented in PoseInference, and PoseInference.

◆ queryParams() [1/2]

virtual void IPoseEstimator::queryParams ( NvMOTQuery &  paramsQuery)
pure virtual

Populate the capabilities reported to the tracker.

Parameters
[out]paramsQueryThe output query parameters.

Implemented in PoseInference, and PoseInference.

◆ queryParams() [2/2]

virtual void IPoseEstimator::queryParams ( NvMOTQuery &  paramsQuery)
pure virtual

Populate the capabilities reported to the tracker.

Parameters
[out]paramsQueryThe output query parameters.

Implemented in PoseInference, and PoseInference.

◆ removeStream() [1/2]

virtual void IPoseEstimator::removeStream ( const NvMOTStreamId  videoStreamId)
pure virtual

Remove a stream from the pose estimator.

Parameters
[in]videoStreamIdVideo stream id.

Implemented in PoseInference, and PoseInference.

◆ removeStream() [2/2]

virtual void IPoseEstimator::removeStream ( const NvMOTStreamId  videoStreamId)
pure virtual

Remove a stream from the pose estimator.

Parameters
[in]videoStreamIdVideo stream id.

Implemented in PoseInference, and PoseInference.

◆ update() [1/2]

virtual void IPoseEstimator::update ( )
pure virtual

Update the pose estimator after data association.

Implemented in PoseInference, and PoseInference.

◆ update() [2/2]

virtual void IPoseEstimator::update ( )
pure virtual

Update the pose estimator after data association.

Implemented in PoseInference, and PoseInference.


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