Detailed Description

Renders and writes visual tracker debugging output.

Definition at line 43 of file sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

Inheritance diagram for TrackerVizOutput:

Collaboration diagram for TrackerVizOutput:

[legend]

Public Member Functions
	TrackerVizOutput (VisualTracker *visTracker, const std::shared_ptr< NvTrackedObjectManager > targets)
	Construct output rendering for a CUDA visual tracker. More...

	TrackerVizOutput (DCFTracker *dcfTracker, const std::shared_ptr< NvTrackedObjectManager > targets)
	Construct output rendering for a VPI DCF tracker. More...

	~TrackerVizOutput ()
	Destroy the visualization output writer. More...

cv::Mat	make_dist_colors (unsigned nb)
	Generate visually distinct BGR colors by distributing shuffled hues. More...

void	writeBirdsEyeView (const uint outputDirectoryKey=0)
	Render and write a bird's-eye view for each available stream. More...

void	writeAllTrajectoriesOnEveryFrame (const uint outputDirectoryKey=0)
	Draw all current trajectories on each available source frame and write the results. More...

void	writeEachCompletedTrajectory (const uint outputDirectoryKey=0)
	Draw and write each trajectory after its track completes. More...

void	writeSrcImagePatchesWithCorrResp (const uint outputDirectoryKey=0)
	Write source image patches alongside their correlation-response visualizations. More...

void	writeSrcImagePatchesWithMaskFilter (const uint outputDirectoryKey=0)
	Write source image patches alongside their segmentation-filter masks. More...

	TrackerVizOutput (VisualTracker *visTracker, const std::shared_ptr< NvTrackedObjectManager > targets)
	Construct output rendering for a CUDA visual tracker. More...

	TrackerVizOutput (DCFTracker *dcfTracker, const std::shared_ptr< NvTrackedObjectManager > targets)
	Construct output rendering for a VPI DCF tracker. More...

	~TrackerVizOutput ()
	Destroy the visualization output writer. More...

cv::Mat	make_dist_colors (unsigned nb)
	Generate visually distinct BGR colors by distributing shuffled hues. More...

void	writeBirdsEyeView (const uint outputDirectoryKey=0)
	Render and write a bird's-eye view for each available stream. More...

void	writeAllTrajectoriesOnEveryFrame (const uint outputDirectoryKey=0)
	Draw all current trajectories on each available source frame and write the results. More...

void	writeEachCompletedTrajectory (const uint outputDirectoryKey=0)
	Draw and write each trajectory after its track completes. More...

void	writeSrcImagePatchesWithCorrResp (const uint outputDirectoryKey=0)
	Write source image patches alongside their correlation-response visualizations. More...

void	writeSrcImagePatchesWithMaskFilter (const uint outputDirectoryKey=0)
	Write source image patches alongside their segmentation-filter masks. More...

bool	isImageOutputEnabled (const uint outputDirectoryKey=0)
	Report whether image output is enabled for a directory key. More...

bool	isImageOutputEnabled (const uint outputDirectoryKey=0)
	Report whether image output is enabled for a directory key. More...

bool	isVideoOutputEnabled ()
	Report whether video output is enabled. More...

bool	isVideoOutputEnabled ()
	Report whether video output is enabled. More...

bool	enableVideoOutput (const std::string outputVideoDirectory, const std::string outputVideoName, cv::Size videoSize, const double fps=5.f)
	Create an AVI writer and enable video output. More...

bool	enableVideoOutput (const std::string outputVideoDirectory, const std::string outputVideoName, cv::Size videoSize, const double fps=5.f)
	Create an AVI writer and enable video output. More...

void	enableImageOutput (const std::string outputImageDirectory, const int compressionLevel=95, const uint outputDirectoryKey=0)
	Configure a keyed directory for JPEG image output. More...

void	enableImageOutput (const std::string outputImageDirectory, const int compressionLevel=95, const uint outputDirectoryKey=0)
	Configure a keyed directory for JPEG image output. More...

void	writeVideoOutput (const cv::Mat &img)
	Append an image to the output video. More...

void	writeVideoOutput (const cv::Mat &img)
	Append an image to the output video. More...

void	writeImageOutput (const cv::Mat &img, const std::string imgFilename, const uint outputDirectoryKey=0, const std::string subFolderName="")
	Write an image to a keyed output directory. More...

void	writeImageOutput (const cv::Mat &img, const std::string imgFilename, const uint outputDirectoryKey=0, const std::string subFolderName="")
	Write an image to a keyed output directory. More...

Protected Member Functions
void	writeSrcImagePatchesWithCorrResp_perTarget (const NvMOTStreamId streamId_, const int trackerInd, const uint outputDirectoryKey, cv::Mat srcImgPatch, cv::Mat corrRespMat)
	Buffer a source patch paired with its correlation-response image for one target. More...

void	writeSrcImagePatchesWithMaskFilter_perTarget (const NvMOTStreamId streamId_, const int trackerInd, const uint outputDirectoryKey, cv::Mat srcImgPatch, cv::Mat maskMat)
	Buffer a source patch paired with its segmentation-filter mask for one target. More...

void	addToWriteBuffer (const std::shared_ptr< NvTrackedObject > target, const cv::Mat &img, const uint frameNum, const uint outputDirectoryKey, const std::string subFolderName, const std::string fileNamePrefix="")
	Add an image to a target's pending output buffer. More...

void	writeFromWriteBuffer (const std::shared_ptr< NvTrackedObject > target)
	Write and clear pending images after a target receives a persistent ID. More...

cv::Mat	prepareCorrelationResponseImg (const NvMOTStreamId streamId_, const uint trackerInd, std::vector< cv::Mat > &corrRespMats, const cv::Size dstSize, const bool showPeakPoint=true, const bool showDetBboxes=true)
	Prepare a color-mapped correlation response with optional target and detection markers. More...

cv::Mat	prepareMaskFilterImg (const NvMOTStreamId streamId_, const uint trackerInd, std::vector< cv::Mat > &maskMats, const cv::Size dstSize)
	Prepare an 8-bit BGR visualization of a target's segmentation-filter mask. More...

void	drawFullTrajectory (const uint64_t targetID, const std::list< TargetInfo > &trajectory, cv::Mat &frame)
	Draw a target's complete trajectory on a frame. More...

void	drawRecentTrajectory (const std::shared_ptr< NvTrackedObject > tracker, cv::Mat &frame, const bool showCurrentDetBbox=true, const bool showCurrentTargetBbox=true, const bool showPrediction=false, const bool showPredictionUncertainty=false, const bool showTrajectoryOnFeet=false)
	Draw a target's recent trajectory, current boxes, and optional predictions on a frame. More...

void	drawUncertaintyEllipse (cv::Mat &frame, const cv::Mat P, const cv::Mat Pcov, cv::Scalar color, float scale=1.f)
	Draw a covariance-derived uncertainty ellipse around a predicted position. More...

void	drawProjectedTargetModel (const std::shared_ptr< NvTrackedObject > target, cv::Mat &frame)
	Draw a target's projected human model on the camera frame. More...

void	drawProjectedPeerTrajectory (const PeerTargetDB &peerTargetDB, cv::Mat &frame, const NvMOTStreamId &videoStreamID, const int &frameNum, const std::map< TargetId, cv::Scalar > &targetColorList)
	Draw peer-camera target boxes and trajectories projected into the current camera frame. More...

void	writeSrcImagePatchesWithCorrResp_perTarget (const NvMOTStreamId streamId_, const int trackerInd, const uint outputDirectoryKey, cv::Mat srcImgPatch, cv::Mat corrRespMat)
	Buffer a source patch paired with its correlation-response image for one target. More...

void	writeSrcImagePatchesWithMaskFilter_perTarget (const NvMOTStreamId streamId_, const int trackerInd, const uint outputDirectoryKey, cv::Mat srcImgPatch, cv::Mat maskMat)
	Buffer a source patch paired with its segmentation-filter mask for one target. More...

void	addToWriteBuffer (const std::shared_ptr< NvTrackedObject > target, const cv::Mat &img, const uint frameNum, const uint outputDirectoryKey, const std::string subFolderName, const std::string fileNamePrefix="")
	Add an image to a target's pending output buffer. More...

void	writeFromWriteBuffer (const std::shared_ptr< NvTrackedObject > target)
	Write and clear pending images after a target receives a persistent ID. More...

cv::Mat	prepareCorrelationResponseImg (const NvMOTStreamId streamId_, const uint trackerInd, std::vector< cv::Mat > &corrRespMats, const cv::Size dstSize, const bool showPeakPoint=true, const bool showDetBboxes=true)
	Prepare a color-mapped correlation response with optional target and detection markers. More...

cv::Mat	prepareMaskFilterImg (const NvMOTStreamId streamId_, const uint trackerInd, std::vector< cv::Mat > &maskMats, const cv::Size dstSize)
	Prepare an 8-bit BGR visualization of a target's segmentation-filter mask. More...

void	drawFullTrajectory (const uint64_t targetID, const std::list< TargetInfo > &trajectory, cv::Mat &frame)
	Draw a target's complete trajectory on a frame. More...

void	drawRecentTrajectory (const std::shared_ptr< NvTrackedObject > tracker, cv::Mat &frame, const bool showCurrentDetBbox=true, const bool showCurrentTargetBbox=true, const bool showPrediction=false, const bool showPredictionUncertainty=false, const bool showTrajectoryOnFeet=false)
	Draw a target's recent trajectory, current boxes, and optional predictions on a frame. More...

void	drawUncertaintyEllipse (cv::Mat &frame, const cv::Mat P, const cv::Mat Pcov, cv::Scalar color, float scale=1.f)
	Draw a covariance-derived uncertainty ellipse around a predicted position. More...

void	drawProjectedTargetModel (const std::shared_ptr< NvTrackedObject > target, cv::Mat &frame)
	Draw a target's projected human model on the camera frame. More...

void	drawProjectedPeerTrajectory (const PeerTargetDB &peerTargetDB, cv::Mat &frame, const NvMOTStreamId &videoStreamID, const int &frameNum, const std::map< TargetId, cv::Scalar > &targetColorList)
	Draw peer-camera target boxes and trajectories projected into the current camera frame. More...

Protected Attributes
VisualTracker *	m_visualTracker = NULL
	CUDA visual tracker that supplies visualization data. More...

DCFTracker *	m_dcfTracker = NULL
	VPI DCF tracker that supplies visualization data. More...

const std::shared_ptr< NvTrackedObjectManager >	m_targets
	Manager that supplies tracked objects. More...

const int	m_displayWidth = 1920
	Display/screen width. More...

const int	m_srcImgWH = 240
	Visualization width and height for source image patches. More...

const int	m_featImgWH = 160
	Visualization width and height for feature images. More...

std::map< uint32_t, std::deque< WriteBuffer_t > >	m_writeBuffer
	Buffered outputs grouped by target index. More...

std::map< NvMOTStreamId, uint64_t >	m_lastFrameMatchedInTrajectoryArchive
	The target ID of the latest trajectory archived. More...

std::map< NvMOTStreamId, cv::Mat >	m_lastFrames
	The last source frame kept. More...

cv::VideoWriter *	mOutputVideo
	Video writer used when video output is enabled. More...

bool	mIsVideoOutputEnabled
	Indicates whether video output is enabled. More...

std::vector< int >	compression_params
	OpenCV image-encoding parameters. More...

std::unordered_map< uint, std::string >	mOutputImageDirectory
	Image output directories keyed by caller-defined values. More...

std::unordered_map< uint, bool >	mIsImageOutputEnabled
	Image-output enabled state for each directory key. More...

Constructor & Destructor Documentation

◆ TrackerVizOutput() [1/4]

TrackerVizOutput::TrackerVizOutput	(	VisualTracker *	visTracker,
		const std::shared_ptr< NvTrackedObjectManager >	targets
	)

inline

Construct output rendering for a CUDA visual tracker.

Parameters

[in]	visTracker	The CUDA visual tracker that supplies visualization data.
[in]	targets	The manager that supplies tracked objects.

Definition at line 111 of file sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

◆ TrackerVizOutput() [2/4]

TrackerVizOutput::TrackerVizOutput	(	DCFTracker *	dcfTracker,
		const std::shared_ptr< NvTrackedObjectManager >	targets
	)

inline

Construct output rendering for a VPI DCF tracker.

Parameters

[in]	dcfTracker	The VPI DCF tracker that supplies visualization data.
[in]	targets	The manager that supplies tracked objects.

Definition at line 121 of file sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

◆ ~TrackerVizOutput() [1/2]

TrackerVizOutput::~TrackerVizOutput ( )

inline

Destroy the visualization output writer.

Definition at line 127 of file sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

◆ TrackerVizOutput() [3/4]

TrackerVizOutput::TrackerVizOutput	(	VisualTracker *	visTracker,
		const std::shared_ptr< NvTrackedObjectManager >	targets
	)

inline

Construct output rendering for a CUDA visual tracker.

Parameters

[in]	visTracker	The CUDA visual tracker that supplies visualization data.
[in]	targets	The manager that supplies tracked objects.

Definition at line 111 of file 9.2/sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

◆ TrackerVizOutput() [4/4]

TrackerVizOutput::TrackerVizOutput	(	DCFTracker *	dcfTracker,
		const std::shared_ptr< NvTrackedObjectManager >	targets
	)

inline

Construct output rendering for a VPI DCF tracker.

Parameters

[in]	dcfTracker	The VPI DCF tracker that supplies visualization data.
[in]	targets	The manager that supplies tracked objects.

Definition at line 121 of file 9.2/sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

◆ ~TrackerVizOutput() [2/2]

TrackerVizOutput::~TrackerVizOutput ( )

inline

Destroy the visualization output writer.

Definition at line 127 of file 9.2/sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

Member Function Documentation

◆ addToWriteBuffer() [1/2]

void TrackerVizOutput::addToWriteBuffer	(	const std::shared_ptr< NvTrackedObject >	target,
		const cv::Mat &	img,
		const uint	frameNum,
		const uint	outputDirectoryKey,
		const std::string	subFolderName,
		const std::string	fileNamePrefix = `""`
	)

inlineprotected

Add an image to a target's pending output buffer.

Parameters

[in]	target	The tracked object that owns the output.
[in]	img	The image to buffer.
[in]	frameNum	The associated frame number.
[in]	outputDirectoryKey	The output directory selector.
[in]	subFolderName	The output subfolder name.
[in]	fileNamePrefix	The output filename prefix.

Clear any existing ones.

Definition at line 751 of file sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References m_writeBuffer, and absl::move().

Referenced by writeSrcImagePatchesWithCorrResp_perTarget(), and writeSrcImagePatchesWithMaskFilter_perTarget().

◆ addToWriteBuffer() [2/2]

void TrackerVizOutput::addToWriteBuffer	(	const std::shared_ptr< NvTrackedObject >	target,
		const cv::Mat &	img,
		const uint	frameNum,
		const uint	outputDirectoryKey,
		const std::string	subFolderName,
		const std::string	fileNamePrefix = `""`
	)

inlineprotected

Add an image to a target's pending output buffer.

Parameters

[in]	target	The tracked object that owns the output.
[in]	img	The image to buffer.
[in]	frameNum	The associated frame number.
[in]	outputDirectoryKey	The output directory selector.
[in]	subFolderName	The output subfolder name.
[in]	fileNamePrefix	The output filename prefix.

Clear any existing ones.

Definition at line 751 of file 9.2/sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References m_writeBuffer, and absl::move().

◆ drawFullTrajectory() [1/2]

void TrackerVizOutput::drawFullTrajectory	(	const uint64_t	targetID,
		const std::list< TargetInfo > &	trajectory,
		cv::Mat &	frame
	)

inlineprotected

Draw a target's complete trajectory on a frame.

Parameters

[in]	targetID	The persistent target identifier.
[in]	trajectory	The complete target trajectory.
[in,out]	frame	The frame updated with the rendered trajectory.

Param to be configured.

Show target ID.

Definition at line 971 of file sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References ACTIVE, max(), and PROJECTED.

Referenced by writeEachCompletedTrajectory().

◆ drawFullTrajectory() [2/2]

void TrackerVizOutput::drawFullTrajectory	(	const uint64_t	targetID,
		const std::list< TargetInfo > &	trajectory,
		cv::Mat &	frame
	)

inlineprotected

Draw a target's complete trajectory on a frame.

Parameters

[in]	targetID	The persistent target identifier.
[in]	trajectory	The complete target trajectory.
[in,out]	frame	The frame updated with the rendered trajectory.

Param to be configured.

Show target ID.

Definition at line 971 of file 9.2/sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References ACTIVE, max(), and PROJECTED.

◆ drawProjectedPeerTrajectory() [1/2]

void TrackerVizOutput::drawProjectedPeerTrajectory	(	const PeerTargetDB &	peerTargetDB,
		cv::Mat &	frame,
		const NvMOTStreamId &	videoStreamID,
		const int &	frameNum,
		const std::map< TargetId, cv::Scalar > &	targetColorList
	)

inlineprotected

Draw peer-camera target boxes and trajectories projected into the current camera frame.

Parameters

[in]	peerTargetDB	The peer-camera target database.
[in,out]	frame	The camera frame updated with peer trajectories.
[in]	videoStreamID	The current video stream identifier.
[in]	frameNum	The current frame number.
[in]	targetColorList	Colors already assigned to local target identifiers.

Skip peers whose last frame is more than two frames behind.

Set target trajectory color.

Not the first point.

Definition at line 1354 of file sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References PeerTarget::frameNum, m_targets, make_dist_colors(), PeerTarget::nextState, and PeerTarget::tracklet.

Referenced by writeAllTrajectoriesOnEveryFrame().

◆ drawProjectedPeerTrajectory() [2/2]

void TrackerVizOutput::drawProjectedPeerTrajectory	(	const PeerTargetDB &	peerTargetDB,
		cv::Mat &	frame,
		const NvMOTStreamId &	videoStreamID,
		const int &	frameNum,
		const std::map< TargetId, cv::Scalar > &	targetColorList
	)

inlineprotected

Draw peer-camera target boxes and trajectories projected into the current camera frame.

Parameters

[in]	peerTargetDB	The peer-camera target database.
[in,out]	frame	The camera frame updated with peer trajectories.
[in]	videoStreamID	The current video stream identifier.
[in]	frameNum	The current frame number.
[in]	targetColorList	Colors already assigned to local target identifiers.

Skip peers whose last frame is more than two frames behind.

Set target trajectory color.

Not the first point.

Definition at line 1354 of file 9.2/sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References PeerTarget::frameNum, m_targets, make_dist_colors(), PeerTarget::nextState, and PeerTarget::tracklet.

◆ drawProjectedTargetModel() [1/2]

void TrackerVizOutput::drawProjectedTargetModel	(	const std::shared_ptr< NvTrackedObject >	target,
		cv::Mat &	frame
	)

inlineprotected

Draw a target's projected human model on the camera frame.

Parameters

[in]	target	The tracked object whose model is drawn.
[in,out]	frame	The camera frame updated with the projected model.

If the last target state is not Active, then don't draw.

Draw the cylinderic human model.

Draw the feet location.

Definition at line 1292 of file sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References ACTIVE, CV_FUNCS::convexHull(), TargetInfo::convexHull, TargetInfo::loc, TargetInfo::pt_img_feet, and TENTATIVE.

Referenced by writeAllTrajectoriesOnEveryFrame().

◆ drawProjectedTargetModel() [2/2]

void TrackerVizOutput::drawProjectedTargetModel	(	const std::shared_ptr< NvTrackedObject >	target,
		cv::Mat &	frame
	)

inlineprotected

Draw a target's projected human model on the camera frame.

Parameters

[in]	target	The tracked object whose model is drawn.
[in,out]	frame	The camera frame updated with the projected model.

If the last target state is not Active, then don't draw.

Draw the cylinderic human model.

Draw the feet location.

Definition at line 1292 of file 9.2/sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References ACTIVE, CV_FUNCS::convexHull(), TargetInfo::convexHull, TargetInfo::loc, TargetInfo::pt_img_feet, and TENTATIVE.

◆ drawRecentTrajectory() [1/2]

void TrackerVizOutput::drawRecentTrajectory	(	const std::shared_ptr< NvTrackedObject >	tracker,
		cv::Mat &	frame,
		const bool	showCurrentDetBbox = `true`,
		const bool	showCurrentTargetBbox = `true`,
		const bool	showPrediction = `false`,
		const bool	showPredictionUncertainty = `false`,
		const bool	showTrajectoryOnFeet = `false`
	)

inlineprotected

Draw a target's recent trajectory, current boxes, and optional predictions on a frame.

Parameters

[in]	tracker	The tracked object whose trajectory is drawn.
[in,out]	frame	The frame updated with the rendered trajectory.
[in]	showCurrentDetBbox	Whether to draw the current detector bounding box.
[in]	showCurrentTargetBbox	Whether to draw the current target bounding box.
[in]	showPrediction	Whether to draw predicted target positions.
[in]	showPredictionUncertainty	Whether to draw prediction uncertainty.
[in]	showTrajectoryOnFeet	Whether to anchor trajectory points at projected feet locations.

Param to be configured.

If the last target state is not Active, then don't draw.

Show target ID.

Consider only the latest N points.

Feet location.

Bbox center.

Draw original detection bbox if there is.

Draw current bbox.

Draw the bbox center.

Time intervals for prediction.

The drawn radius is a fixed stand-in for uncertainty.

Two ellipses to show the prediction and uncertainty.

Draw the projected circle.

Time intervals for prediction.

Two ellipses to show the prediction and uncertainty.

Definition at line 1038 of file sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References ACTIVE, _NvMOTObjToTrack::bbox, CV_FUNCS::constructRect2f(), drawUncertaintyEllipse(), _NvMOTRect::height, max(), TENTATIVE, _NvMOTRect::width, _NvMOTRect::x, and _NvMOTRect::y.

Referenced by writeAllTrajectoriesOnEveryFrame().

◆ drawRecentTrajectory() [2/2]

void TrackerVizOutput::drawRecentTrajectory	(	const std::shared_ptr< NvTrackedObject >	tracker,
		cv::Mat &	frame,
		const bool	showCurrentDetBbox = `true`,
		const bool	showCurrentTargetBbox = `true`,
		const bool	showPrediction = `false`,
		const bool	showPredictionUncertainty = `false`,
		const bool	showTrajectoryOnFeet = `false`
	)

inlineprotected

Draw a target's recent trajectory, current boxes, and optional predictions on a frame.

Parameters

[in]	tracker	The tracked object whose trajectory is drawn.
[in,out]	frame	The frame updated with the rendered trajectory.
[in]	showCurrentDetBbox	Whether to draw the current detector bounding box.
[in]	showCurrentTargetBbox	Whether to draw the current target bounding box.
[in]	showPrediction	Whether to draw predicted target positions.
[in]	showPredictionUncertainty	Whether to draw prediction uncertainty.
[in]	showTrajectoryOnFeet	Whether to anchor trajectory points at projected feet locations.

Param to be configured.

If the last target state is not Active, then don't draw.

Show target ID.

Consider only the latest N points.

Feet location.

Bbox center.

Draw original detection bbox if there is.

Draw current bbox.

Draw the bbox center.

Time intervals for prediction.

The drawn radius is a fixed stand-in for uncertainty.

Two ellipses to show the prediction and uncertainty.

Draw the projected circle.

Time intervals for prediction.

Two ellipses to show the prediction and uncertainty.

Definition at line 1038 of file 9.2/sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References ACTIVE, _NvMOTObjToTrack::bbox, CV_FUNCS::constructRect2f(), drawUncertaintyEllipse(), _NvMOTRect::height, max(), TENTATIVE, _NvMOTRect::width, _NvMOTRect::x, and _NvMOTRect::y.

◆ drawUncertaintyEllipse() [1/2]

void TrackerVizOutput::drawUncertaintyEllipse	(	cv::Mat &	frame,
		const cv::Mat	P,
		const cv::Mat	Pcov,
		cv::Scalar	color,
		float	scale = `1.f`
	)

inlineprotected

Draw a covariance-derived uncertainty ellipse around a predicted position.

Parameters

[in,out]	frame	The frame updated with the rendered ellipse.
[in]	P	The predicted position vector.
[in]	Pcov	The predicted position covariance matrix.
[in]	color	The ellipse color.
[in]	scale	The ellipse-size scale factor.

Compute the vanishing correlation coefficient.

Compute the principal semi-diameters of the ellipse.

Compute the angle between the x1 axis and the semi-diameter of length p1.

Convert from Radian to Degree.

Drawing the ellipse.

Definition at line 1247 of file 9.2/sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References half_float::atan(), half_float::cos(), half_float::pow(), half_float::sin(), half_float::detail::sqrt(), and width.

◆ drawUncertaintyEllipse() [2/2]

void TrackerVizOutput::drawUncertaintyEllipse	(	cv::Mat &	frame,
		const cv::Mat	P,
		const cv::Mat	Pcov,
		cv::Scalar	color,
		float	scale = `1.f`
	)

inlineprotected

Draw a covariance-derived uncertainty ellipse around a predicted position.

Parameters

[in,out]	frame	The frame updated with the rendered ellipse.
[in]	P	The predicted position vector.
[in]	Pcov	The predicted position covariance matrix.
[in]	color	The ellipse color.
[in]	scale	The ellipse-size scale factor.

Compute the vanishing correlation coefficient.

Compute the principal semi-diameters of the ellipse.

Compute the angle between the x1 axis and the semi-diameter of length p1.

Convert from Radian to Degree.

Drawing the ellipse.

Definition at line 1247 of file sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References half_float::atan(), half_float::cos(), half_float::pow(), half_float::sin(), half_float::detail::sqrt(), and width.

Referenced by drawRecentTrajectory().

◆ enableImageOutput() [1/2]

void FrameOutput::enableImageOutput	(	const std::string	outputImageDirectory,
		const int	compressionLevel = `95`,
		const uint	outputDirectoryKey = `0`
	)

inlineinherited

Configure a keyed directory for JPEG image output.

Parameters

[in]	outputImageDirectory	Directory in which images are written.
[in]	compressionLevel	JPEG quality level.
[in]	outputDirectoryKey	Key used to select the output directory.

Set the JPEG image quality.

Definition at line 116 of file sources/libs/nvmultiobjecttracker/include/modules/utility/FrameOutput.hpp.

References FrameOutput::compression_params, FrameOutput::mIsImageOutputEnabled, and FrameOutput::mOutputImageDirectory.

◆ enableImageOutput() [2/2]

void FrameOutput::enableImageOutput	(	const std::string	outputImageDirectory,
		const int	compressionLevel = `95`,
		const uint	outputDirectoryKey = `0`
	)

inlineinherited

Configure a keyed directory for JPEG image output.

Parameters

[in]	outputImageDirectory	Directory in which images are written.
[in]	compressionLevel	JPEG quality level.
[in]	outputDirectoryKey	Key used to select the output directory.

Set the JPEG image quality.

Definition at line 116 of file 9.2/sources/libs/nvmultiobjecttracker/include/modules/utility/FrameOutput.hpp.

References FrameOutput::compression_params, FrameOutput::mIsImageOutputEnabled, and FrameOutput::mOutputImageDirectory.

◆ enableVideoOutput() [1/2]

bool FrameOutput::enableVideoOutput	(	const std::string	outputVideoDirectory,
		const std::string	outputVideoName,
		cv::Size	videoSize,
		const double	fps = `5.f`
	)

inlineinherited

Create an AVI writer and enable video output.

Parameters

[in]	outputVideoDirectory	Directory in which the video is written.
[in]	outputVideoName	Base name of the output video file.
[in]	videoSize	Output video frame size.
[in]	fps	Output video frame rate.

Returns: True when the output video is opened successfully.

Definition at line 78 of file sources/libs/nvmultiobjecttracker/include/modules/utility/FrameOutput.hpp.

References FrameOutput::mIsVideoOutputEnabled, and FrameOutput::mOutputVideo.

◆ enableVideoOutput() [2/2]

bool FrameOutput::enableVideoOutput	(	const std::string	outputVideoDirectory,
		const std::string	outputVideoName,
		cv::Size	videoSize,
		const double	fps = `5.f`
	)

inlineinherited

Create an AVI writer and enable video output.

Parameters

[in]	outputVideoDirectory	Directory in which the video is written.
[in]	outputVideoName	Base name of the output video file.
[in]	videoSize	Output video frame size.
[in]	fps	Output video frame rate.

Returns: True when the output video is opened successfully.

Definition at line 78 of file 9.2/sources/libs/nvmultiobjecttracker/include/modules/utility/FrameOutput.hpp.

References FrameOutput::mIsVideoOutputEnabled, and FrameOutput::mOutputVideo.

◆ isImageOutputEnabled() [1/2]

bool FrameOutput::isImageOutputEnabled ( const uint outputDirectoryKey = 0 )

inlineinherited

Report whether image output is enabled for a directory key.

Parameters

[in] outputDirectoryKey Output directory key to query.

Definition at line 55 of file sources/libs/nvmultiobjecttracker/include/modules/utility/FrameOutput.hpp.

References FrameOutput::mIsImageOutputEnabled.

Referenced by FrameOutput::writeImageOutput().

◆ isImageOutputEnabled() [2/2]

bool FrameOutput::isImageOutputEnabled ( const uint outputDirectoryKey = 0 )

inlineinherited

Report whether image output is enabled for a directory key.

Parameters

[in] outputDirectoryKey Output directory key to query.

Definition at line 55 of file 9.2/sources/libs/nvmultiobjecttracker/include/modules/utility/FrameOutput.hpp.

References FrameOutput::mIsImageOutputEnabled.

◆ isVideoOutputEnabled() [1/2]

bool FrameOutput::isVideoOutputEnabled ( )

inlineinherited

Report whether video output is enabled.

Definition at line 68 of file sources/libs/nvmultiobjecttracker/include/modules/utility/FrameOutput.hpp.

References FrameOutput::mIsVideoOutputEnabled.

◆ isVideoOutputEnabled() [2/2]

bool FrameOutput::isVideoOutputEnabled ( )

inlineinherited

Report whether video output is enabled.

Definition at line 68 of file 9.2/sources/libs/nvmultiobjecttracker/include/modules/utility/FrameOutput.hpp.

References FrameOutput::mIsVideoOutputEnabled.

◆ make_dist_colors() [1/2]

cv::Mat TrackerVizOutput::make_dist_colors ( unsigned nb )

inline

Generate visually distinct BGR colors by distributing shuffled hues.

Parameters

[in] nb The number of colors to generate.

Returns: A single-row BGR matrix containing the generated colors.

Compute step to cover the whole hue range in OpenCV.

Generate hue values.

Shuffle the hue values randomly.

Build the HSV matrix with shuffled hues. High saturation and value for vivid colors.

Definition at line 134 of file sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

Referenced by drawProjectedPeerTrajectory().

◆ make_dist_colors() [2/2]

cv::Mat TrackerVizOutput::make_dist_colors ( unsigned nb )

inline

Generate visually distinct BGR colors by distributing shuffled hues.

Parameters

[in] nb The number of colors to generate.

Returns: A single-row BGR matrix containing the generated colors.

Compute step to cover the whole hue range in OpenCV.

Generate hue values.

Shuffle the hue values randomly.

Build the HSV matrix with shuffled hues. High saturation and value for vivid colors.

Definition at line 134 of file 9.2/sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

◆ prepareCorrelationResponseImg() [1/2]

cv::Mat TrackerVizOutput::prepareCorrelationResponseImg	(	const NvMOTStreamId	streamId_,
		const uint	trackerInd,
		std::vector< cv::Mat > &	corrRespMats,
		const cv::Size	dstSize,
		const bool	showPeakPoint = `true`,
		const bool	showDetBboxes = `true`
	)

inlineprotected

Prepare a color-mapped correlation response with optional target and detection markers.

Parameters

[in]	streamId_	The video stream identifier.
[in]	trackerInd	The target's tracker index.
[in]	corrRespMats	The correlation-response matrices for the tracker batch.
[in]	dstSize	The requested output image size.
[in]	showPeakPoint	Whether to draw the localized peak.
[in]	showDetBboxes	Whether to draw nearby detector bounding-box centers.

Returns: The prepared BGR correlation-response image.

Normalize to a high-enough magnitude for better visualization.

Wrap around with the circular shifting.

Draw the peak point.

Draw the detector bbox points within the search region.

Convert the center point of the detector bbox into the tracker corr resp's coordinate system.

Definition at line 809 of file sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References CV_FUNCS::circshift(), EMPTY, m_dcfTracker, DCFTracker::m_detectorBboxes_h, VisualTracker::m_detectorBboxes_h, DCFTracker::m_detectorBboxes_size_h, VisualTracker::m_detectorBboxes_size_h, VisualTracker::m_targetDisplacement2D_h, m_targets, m_visualTracker, normalize(), pydeepstreamdoc::tensor::TensorDoc::size, width, detectorBbox_t::x, and detectorBbox_t::y.

Referenced by writeSrcImagePatchesWithCorrResp().

◆ prepareCorrelationResponseImg() [2/2]

cv::Mat TrackerVizOutput::prepareCorrelationResponseImg	(	const NvMOTStreamId	streamId_,
		const uint	trackerInd,
		std::vector< cv::Mat > &	corrRespMats,
		const cv::Size	dstSize,
		const bool	showPeakPoint = `true`,
		const bool	showDetBboxes = `true`
	)

inlineprotected

Prepare a color-mapped correlation response with optional target and detection markers.

Parameters

[in]	streamId_	The video stream identifier.
[in]	trackerInd	The target's tracker index.
[in]	corrRespMats	The correlation-response matrices for the tracker batch.
[in]	dstSize	The requested output image size.
[in]	showPeakPoint	Whether to draw the localized peak.
[in]	showDetBboxes	Whether to draw nearby detector bounding-box centers.

Returns: The prepared BGR correlation-response image.

Normalize to a high-enough magnitude for better visualization.

Wrap around with the circular shifting.

Draw the peak point.

Draw the detector bbox points within the search region.

Convert the center point of the detector bbox into the tracker corr resp's coordinate system.

Definition at line 809 of file 9.2/sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References CV_FUNCS::circshift(), EMPTY, m_dcfTracker, DCFTracker::m_detectorBboxes_h, VisualTracker::m_detectorBboxes_h, DCFTracker::m_detectorBboxes_size_h, VisualTracker::m_detectorBboxes_size_h, VisualTracker::m_targetDisplacement2D_h, m_targets, m_visualTracker, normalize(), pydeepstreamdoc::tensor::TensorDoc::size, width, detectorBbox_t::x, and detectorBbox_t::y.

◆ prepareMaskFilterImg() [1/2]

cv::Mat TrackerVizOutput::prepareMaskFilterImg	(	const NvMOTStreamId	streamId_,
		const uint	trackerInd,
		std::vector< cv::Mat > &	maskMats,
		const cv::Size	dstSize
	)

inlineprotected

Prepare an 8-bit BGR visualization of a target's segmentation-filter mask.

Parameters

[in]	streamId_	The video stream identifier.
[in]	trackerInd	The target's tracker index.
[in]	maskMats	The segmentation-filter masks for the tracker batch.
[in]	dstSize	The requested output image size.

Returns: The prepared BGR mask image.

Scale the mask from [0.0, 1.0] into 8-bit values.

Definition at line 931 of file sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References EMPTY, m_targets, pydeepstreamdoc::tensor::TensorDoc::size, and width.

Referenced by writeSrcImagePatchesWithMaskFilter().

◆ prepareMaskFilterImg() [2/2]

cv::Mat TrackerVizOutput::prepareMaskFilterImg	(	const NvMOTStreamId	streamId_,
		const uint	trackerInd,
		std::vector< cv::Mat > &	maskMats,
		const cv::Size	dstSize
	)

inlineprotected

Prepare an 8-bit BGR visualization of a target's segmentation-filter mask.

Parameters

[in]	streamId_	The video stream identifier.
[in]	trackerInd	The target's tracker index.
[in]	maskMats	The segmentation-filter masks for the tracker batch.
[in]	dstSize	The requested output image size.

Returns: The prepared BGR mask image.

Scale the mask from [0.0, 1.0] into 8-bit values.

Definition at line 931 of file 9.2/sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References EMPTY, m_targets, pydeepstreamdoc::tensor::TensorDoc::size, and width.

◆ writeAllTrajectoriesOnEveryFrame() [1/2]

void TrackerVizOutput::writeAllTrajectoriesOnEveryFrame ( const uint outputDirectoryKey = 0 )

inline

Draw all current trajectories on each available source frame and write the results.

Parameters

[in] outputDirectoryKey The output directory selector.

Get source image frames available at Host (made available only when m_params.m_doTrajectoryAnalysis is enabled). {videoStreamID, src image frame}.

Get source image patches to Host.

Write all the recent trajectories on the source frame.

Show frame number.

Draw recent trajectories.

Draw projected human models on the camera frame.

Definition at line 405 of file 9.2/sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References CLOG, drawProjectedPeerTrajectory(), drawProjectedTargetModel(), drawRecentTrajectory(), m_dcfTracker, m_targets, m_visualTracker, and FrameOutput::writeImageOutput().

◆ writeAllTrajectoriesOnEveryFrame() [2/2]

void TrackerVizOutput::writeAllTrajectoriesOnEveryFrame ( const uint outputDirectoryKey = 0 )

inline

Draw all current trajectories on each available source frame and write the results.

Parameters

[in] outputDirectoryKey The output directory selector.

Get source image frames available at Host (made available only when m_params.m_doTrajectoryAnalysis is enabled). {videoStreamID, src image frame}.

Get source image patches to Host.

Write all the recent trajectories on the source frame.

Show frame number.

Draw recent trajectories.

Draw projected human models on the camera frame.

Definition at line 405 of file sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References CLOG, drawProjectedPeerTrajectory(), drawProjectedTargetModel(), drawRecentTrajectory(), m_dcfTracker, m_targets, m_visualTracker, and FrameOutput::writeImageOutput().

◆ writeBirdsEyeView() [1/2]

void TrackerVizOutput::writeBirdsEyeView ( const uint outputDirectoryKey = 0 )

inline

Render and write a bird's-eye view for each available stream.

Parameters

[in] outputDirectoryKey The output directory selector.

{videoStreamID, src image frame}.

{videoStreamID, world map coordinates}.

{videoStreamID, world-to-frame scale}.

Both frame dimensions use the source height.

Define the world map size.

If the last target state is not Active, then don't draw.

Check if a point in world coordinate is within the world map size. If not, adjust world map size to include the new point.

Param to be configured.

Draw the trajectories of each stream on its bird's-eye view frame.

Show frame number.

If the last target state is not Active, then don't draw.

Make the image coordinate axis align with the world coordinate axis in terms of orientations. The origin of the image coordinate is at the top-left, while the world coordinate origin is at the bottom-right.

Draw the bbox center.

Show target label (ID & Age).

Definition at line 163 of file sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References ACTIVE, NvMotDataTypes::Size::height, m_dcfTracker, VisualTracker::m_paramsRt, DCFTracker::m_paramsRt, m_targets, m_visualTracker, max(), VisualTrackerRunTimeParams::srcImageSize, DCFTrackerRunTimeParams::srcImageSize, TENTATIVE, and FrameOutput::writeImageOutput().

◆ writeBirdsEyeView() [2/2]

void TrackerVizOutput::writeBirdsEyeView ( const uint outputDirectoryKey = 0 )

inline

Render and write a bird's-eye view for each available stream.

Parameters

[in] outputDirectoryKey The output directory selector.

{videoStreamID, src image frame}.

{videoStreamID, world map coordinates}.

{videoStreamID, world-to-frame scale}.

Both frame dimensions use the source height.

Define the world map size.

If the last target state is not Active, then don't draw.

Check if a point in world coordinate is within the world map size. If not, adjust world map size to include the new point.

Param to be configured.

Draw the trajectories of each stream on its bird's-eye view frame.

Show frame number.

If the last target state is not Active, then don't draw.

Make the image coordinate axis align with the world coordinate axis in terms of orientations. The origin of the image coordinate is at the top-left, while the world coordinate origin is at the bottom-right.

Draw the bbox center.

Show target label (ID & Age).

Definition at line 163 of file 9.2/sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References ACTIVE, NvMotDataTypes::Size::height, m_dcfTracker, VisualTracker::m_paramsRt, DCFTracker::m_paramsRt, m_targets, m_visualTracker, max(), VisualTrackerRunTimeParams::srcImageSize, DCFTrackerRunTimeParams::srcImageSize, TENTATIVE, and FrameOutput::writeImageOutput().

◆ writeEachCompletedTrajectory() [1/2]

void TrackerVizOutput::writeEachCompletedTrajectory ( const uint outputDirectoryKey = 0 )

inline

Draw and write each trajectory after its track completes.

Parameters

[in] outputDirectoryKey The output directory selector.

Get source image frames available at Host (made available only when m_params.m_doTrajectoryAnalysis is enabled). {videoStreamID, src image frame}.

Get source image patches to Host.

Keep the latest copy of the source frame.

Write a complete trajectory of a target when terminated.

{targetId, Trajectory4Archive}.

Write the processed tracklet for visualization.

Definition at line 500 of file sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References CLOG, drawFullTrajectory(), m_dcfTracker, m_lastFrames, m_targets, m_visualTracker, and FrameOutput::writeImageOutput().

◆ writeEachCompletedTrajectory() [2/2]

void TrackerVizOutput::writeEachCompletedTrajectory ( const uint outputDirectoryKey = 0 )

inline

Draw and write each trajectory after its track completes.

Parameters

[in] outputDirectoryKey The output directory selector.

Get source image frames available at Host (made available only when m_params.m_doTrajectoryAnalysis is enabled). {videoStreamID, src image frame}.

Get source image patches to Host.

Keep the latest copy of the source frame.

Write a complete trajectory of a target when terminated.

{targetId, Trajectory4Archive}.

Write the processed tracklet for visualization.

Definition at line 500 of file 9.2/sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References CLOG, drawFullTrajectory(), m_dcfTracker, m_lastFrames, m_targets, m_visualTracker, and FrameOutput::writeImageOutput().

◆ writeFromWriteBuffer() [1/2]

void TrackerVizOutput::writeFromWriteBuffer ( const std::shared_ptr< NvTrackedObject > target )

inlineprotected

Write and clear pending images after a target receives a persistent ID.

Parameters

[in] target The tracked object whose pending images are written.

FileName: Target_{targetID}_Frame_{FrameNum}.

Out folder name: Stream_{StreamID}.

Definition at line 772 of file 9.2/sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References ACTIVE, INACTIVE, m_writeBuffer, pydeepstreamdoc::tensor::TensorDoc::size, and FrameOutput::writeImageOutput().

◆ writeFromWriteBuffer() [2/2]

void TrackerVizOutput::writeFromWriteBuffer ( const std::shared_ptr< NvTrackedObject > target )

inlineprotected

Write and clear pending images after a target receives a persistent ID.

Parameters

[in] target The tracked object whose pending images are written.

FileName: Target_{targetID}_Frame_{FrameNum}.

Out folder name: Stream_{StreamID}.

Definition at line 772 of file sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References ACTIVE, INACTIVE, m_writeBuffer, pydeepstreamdoc::tensor::TensorDoc::size, and FrameOutput::writeImageOutput().

Referenced by writeSrcImagePatchesWithCorrResp_perTarget(), and writeSrcImagePatchesWithMaskFilter_perTarget().

◆ writeImageOutput() [1/2]

void FrameOutput::writeImageOutput	(	const cv::Mat &	img,
		const std::string	imgFilename,
		const uint	outputDirectoryKey = `0`,
		const std::string	subFolderName = `""`
	)

inlineinherited

Write an image to a keyed output directory.

Parameters

[in]	img	Image to write.
[in]	imgFilename	Base name of the output image file.
[in]	outputDirectoryKey	Key used to select the output directory.
[in]	subFolderName	Optional subdirectory below the keyed output directory.

Create the output directory if it does not exist.

Definition at line 157 of file sources/libs/nvmultiobjecttracker/include/modules/utility/FrameOutput.hpp.

References FrameOutput::compression_params, FrameOutput::isImageOutputEnabled(), FrameOutput::mOutputImageDirectory, and pydeepstreamdoc::tensor::TensorDoc::size.

Referenced by writeAllTrajectoriesOnEveryFrame(), writeBirdsEyeView(), writeEachCompletedTrajectory(), and writeFromWriteBuffer().

◆ writeImageOutput() [2/2]

void FrameOutput::writeImageOutput	(	const cv::Mat &	img,
		const std::string	imgFilename,
		const uint	outputDirectoryKey = `0`,
		const std::string	subFolderName = `""`
	)

inlineinherited

Write an image to a keyed output directory.

Parameters

[in]	img	Image to write.
[in]	imgFilename	Base name of the output image file.
[in]	outputDirectoryKey	Key used to select the output directory.
[in]	subFolderName	Optional subdirectory below the keyed output directory.

Create the output directory if it does not exist.

Definition at line 157 of file 9.2/sources/libs/nvmultiobjecttracker/include/modules/utility/FrameOutput.hpp.

References FrameOutput::compression_params, FrameOutput::isImageOutputEnabled(), FrameOutput::mOutputImageDirectory, and pydeepstreamdoc::tensor::TensorDoc::size.

◆ writeSrcImagePatchesWithCorrResp() [1/2]

void TrackerVizOutput::writeSrcImagePatchesWithCorrResp ( const uint outputDirectoryKey = 0 )

inline

Write source image patches alongside their correlation-response visualizations.

Parameters

[in] outputDirectoryKey The output directory selector.

Get source image patches to Host.

Definition at line 569 of file 9.2/sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References pydeepstreamdoc::tensor::TensorDoc::clone, DCFTracker::m_corrResponses_d, VisualTracker::m_corrResponses_d, m_dcfTracker, VisualTracker::m_featureExtractor, VisualTracker::m_params, DCFTracker::m_params, m_srcImgWH, m_targets, m_visualTracker, VisualTrackerConfigParams::m_vizDebugStreamId, DCFTrackerConfigParams::m_vizDebugStreamId, prepareCorrelationResponseImg(), and writeSrcImagePatchesWithCorrResp_perTarget().

◆ writeSrcImagePatchesWithCorrResp() [2/2]

void TrackerVizOutput::writeSrcImagePatchesWithCorrResp ( const uint outputDirectoryKey = 0 )

inline

Write source image patches alongside their correlation-response visualizations.

Parameters

[in] outputDirectoryKey The output directory selector.

Get source image patches to Host.

Definition at line 569 of file sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References pydeepstreamdoc::tensor::TensorDoc::clone, DCFTracker::m_corrResponses_d, VisualTracker::m_corrResponses_d, m_dcfTracker, VisualTracker::m_featureExtractor, VisualTracker::m_params, DCFTracker::m_params, m_srcImgWH, m_targets, m_visualTracker, VisualTrackerConfigParams::m_vizDebugStreamId, DCFTrackerConfigParams::m_vizDebugStreamId, prepareCorrelationResponseImg(), and writeSrcImagePatchesWithCorrResp_perTarget().

◆ writeSrcImagePatchesWithCorrResp_perTarget() [1/2]

void TrackerVizOutput::writeSrcImagePatchesWithCorrResp_perTarget	(	const NvMOTStreamId	streamId_,
		const int	trackerInd,
		const uint	outputDirectoryKey,
		cv::Mat	srcImgPatch,
		cv::Mat	corrRespMat
	)

inlineprotected

Buffer a source patch paired with its correlation-response image for one target.

Parameters

[in]	streamId_	The video stream identifier.
[in]	trackerInd	The target's tracker index.
[in]	outputDirectoryKey	The output directory selector.
[in]	srcImgPatch	The source image patch.
[in]	corrRespMat	The correlation-response visualization.

Stack the two.

Put it into a buffer first because if target state is TENTATIVE, there is no ID assigned yet.

Write into file if ID is assigned.

Definition at line 658 of file sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References addToWriteBuffer(), VisualTrackerRunTimeParams::colorFormat, EMPTY, VisualTracker::m_paramsRt, m_srcImgWH, m_targets, m_visualTracker, NVBUF_COLOR_FORMAT_RGBA, and writeFromWriteBuffer().

Referenced by writeSrcImagePatchesWithCorrResp().

◆ writeSrcImagePatchesWithCorrResp_perTarget() [2/2]

void TrackerVizOutput::writeSrcImagePatchesWithCorrResp_perTarget	(	const NvMOTStreamId	streamId_,
		const int	trackerInd,
		const uint	outputDirectoryKey,
		cv::Mat	srcImgPatch,
		cv::Mat	corrRespMat
	)

inlineprotected

Buffer a source patch paired with its correlation-response image for one target.

Parameters

[in]	streamId_	The video stream identifier.
[in]	trackerInd	The target's tracker index.
[in]	outputDirectoryKey	The output directory selector.
[in]	srcImgPatch	The source image patch.
[in]	corrRespMat	The correlation-response visualization.

Stack the two.

Put it into a buffer first because if target state is TENTATIVE, there is no ID assigned yet.

Write into file if ID is assigned.

Definition at line 658 of file 9.2/sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References addToWriteBuffer(), VisualTrackerRunTimeParams::colorFormat, EMPTY, VisualTracker::m_paramsRt, m_srcImgWH, m_targets, m_visualTracker, NVBUF_COLOR_FORMAT_RGBA, and writeFromWriteBuffer().

◆ writeSrcImagePatchesWithMaskFilter() [1/2]

void TrackerVizOutput::writeSrcImagePatchesWithMaskFilter ( const uint outputDirectoryKey = 0 )

inline

Write source image patches alongside their segmentation-filter masks.

Parameters

[in] outputDirectoryKey The output directory selector.

Get source image patches to Host.

Definition at line 611 of file 9.2/sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References pydeepstreamdoc::tensor::TensorDoc::clone, VisualTracker::m_featureExtractor, VisualTracker::m_maskingSegWindow4Model, VisualTracker::m_params, m_srcImgWH, m_targets, m_visualTracker, VisualTrackerConfigParams::m_vizDebugStreamId, prepareMaskFilterImg(), and writeSrcImagePatchesWithMaskFilter_perTarget().

◆ writeSrcImagePatchesWithMaskFilter() [2/2]

void TrackerVizOutput::writeSrcImagePatchesWithMaskFilter ( const uint outputDirectoryKey = 0 )

inline

Write source image patches alongside their segmentation-filter masks.

Parameters

[in] outputDirectoryKey The output directory selector.

Get source image patches to Host.

Definition at line 611 of file sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References pydeepstreamdoc::tensor::TensorDoc::clone, VisualTracker::m_featureExtractor, VisualTracker::m_maskingSegWindow4Model, VisualTracker::m_params, m_srcImgWH, m_targets, m_visualTracker, VisualTrackerConfigParams::m_vizDebugStreamId, prepareMaskFilterImg(), and writeSrcImagePatchesWithMaskFilter_perTarget().

◆ writeSrcImagePatchesWithMaskFilter_perTarget() [1/2]

void TrackerVizOutput::writeSrcImagePatchesWithMaskFilter_perTarget	(	const NvMOTStreamId	streamId_,
		const int	trackerInd,
		const uint	outputDirectoryKey,
		cv::Mat	srcImgPatch,
		cv::Mat	maskMat
	)

inlineprotected

Buffer a source patch paired with its segmentation-filter mask for one target.

Parameters

[in]	streamId_	The video stream identifier.
[in]	trackerInd	The target's tracker index.
[in]	outputDirectoryKey	The output directory selector.
[in]	srcImgPatch	The source image patch.
[in]	maskMat	The segmentation-filter mask visualization.

Stack the two.

Put it into a buffer first because if target state is TENTATIVE, there is no ID assigned yet.

Write into file if ID is assigned.

Definition at line 704 of file 9.2/sources/libs/nvmultiobjecttracker/include/modules/VisualTracker/vizOutput.hpp.

References addToWriteBuffer(), VisualTrackerRunTimeParams::colorFormat, EMPTY, VisualTracker::m_paramsRt, m_srcImgWH, m_targets, m_visualTracker, NVBUF_COLOR_FORMAT_RGBA, and writeFromWriteBuffer().

◆ writeSrcImagePatchesWithMaskFilter_perTarget() [2/2]

void TrackerVizOutput::writeSrcImagePatchesWithMaskFilter_perTarget	(	const NvMOTStreamId	streamId_,
		const int	trackerInd,
		const uint	outputDirectoryKey,
		cv::Mat	srcImgPatch,
		cv::Mat	maskMat
	)

inlineprotected

Buffer a source patch paired with its segmentation-filter mask for one target.

Parameters

[in]	streamId_	The video stream identifier.
[in]	trackerInd	The target's tracker index.
[in]	outputDirectoryKey	The output directory selector.
[in]	srcImgPatch	The source image patch.
[in]	maskMat	The segmentation-filter mask visualization.