sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/postprocess/display_bbox3d.h File Reference

Go to the source code of this file.

Data Structures
struct	ClassThreshold
	Structure to hold class label and its confidence threshold. More...

Functions
std::vector< ClassThreshold >	fetch_class_labels_with_thresholds (const std::string &labels_file, PrettyLogger *logger)
	Fetch class labels and thresholds from a labels file. More...
std::vector< std::string >	extract_class_labels (const std::vector< ClassThreshold > &class_thresholds)
	Extract class label strings from ClassThreshold vector. More...
void	attach_bbox3d_to_display (NvDsBatchMeta *batch_meta, NvDsFrameMeta *frame_meta, const torch::Tensor &camera_frames, const torch::Tensor &boxes_3d, const torch::Tensor &labels_3d, const torch::Tensor &instance_ids, const torch::Tensor &scores_3d, int img_width, int img_height, const std::vector< std::string > &class_labels, const std::string &bbox_color_mode, const bool &display_class_labels, const bool &display_instance_ids, const bool &display_class_score, PrettyLogger *logger)
	Attach 3D bounding boxes to display metadata for visualization. More...
void	clamp_point (float *x, float *y, int img_width, int img_height)
	Clamps a point to stay within image boundaries. More...
void	calculate_centroid (const float *points, float *centroid)
	Calculates the centroid of a 3D bounding box from its 2D projection. More...
void	get_color_from_label (const std::string &label, float *red, float *green, float *blue, int num_classes)
	Generates a consistent color based on a class label. More...
void	get_color_from_instance_id (int64_t instance_id, float *red, float *green, float *blue)
	Generates a color based on instance ID for tracking visualization. More...
bool	count_visible_corners (const float *points, int num_corners, int img_width, int img_height)
	Counts visible corners of a 3D bounding box. More...
void	attach_bbox2d_to_objectmeta (NvDsBatchMeta *batch_meta, NvDsFrameMeta *frame_meta, const torch::Tensor &bbox2d, const torch::Tensor &boxes_3d, const torch::Tensor &labels_3d, const torch::Tensor &instance_ids, const torch::Tensor &scores_3d, const std::vector< std::string > &class_labels, const std::string &bbox_color_mode, const bool &display_2d_bbox, PrettyLogger *logger)
	Attach 2D bounding boxes as NvDsObjectMeta to frame metadata. More...

Function Documentation

attach_bbox2d_to_objectmeta()

void attach_bbox2d_to_objectmeta	(	NvDsBatchMeta *	batch_meta,
		NvDsFrameMeta *	frame_meta,
		const torch::Tensor &	bbox2d,
		const torch::Tensor &	boxes_3d,
		const torch::Tensor &	labels_3d,
		const torch::Tensor &	instance_ids,
		const torch::Tensor &	scores_3d,
		const std::vector< std::string > &	class_labels,
		const std::string &	bbox_color_mode,
		const bool &	display_2d_bbox,
		PrettyLogger *	logger
	)

Attach 2D bounding boxes as NvDsObjectMeta to frame metadata.

Parameters

batch_meta	DeepStream batch metadata
frame_meta	DeepStream frame metadata
bbox2d	Tensor containing 2D bounding boxes [N, 4] where 4 = [x, y, w, h]
boxes_3d	Tensor containing 3D bounding box information
labels_3d	Tensor containing class labels
instance_ids	Tensor containing instance IDs for tracking
scores_3d	Tensor containing confidence scores
class_labels	Vector of class label strings
bbox_color_mode	Color mode: "class_id" for class-based or "instance_id" for instance-based colors
display_2d_bbox	Whether to display 2D bounding boxes on frames
logger	Logger for debug and warning messages

attach_bbox3d_to_display()

void attach_bbox3d_to_display	(	NvDsBatchMeta *	batch_meta,
		NvDsFrameMeta *	frame_meta,
		const torch::Tensor &	camera_frames,
		const torch::Tensor &	boxes_3d,
		const torch::Tensor &	labels_3d,
		const torch::Tensor &	instance_ids,
		const torch::Tensor &	scores_3d,
		int	img_width,
		int	img_height,
		const std::vector< std::string > &	class_labels,
		const std::string &	bbox_color_mode,
		const bool &	display_class_labels,
		const bool &	display_instance_ids,
		const bool &	display_class_score,
		PrettyLogger *	logger
	)

Attach 3D bounding boxes to display metadata for visualization.

This function processes 3D bounding box projections and attaches them to the DeepStream display metadata for visualization on the frame.

Parameters

batch_meta	DeepStream batch metadata
frame_meta	DeepStream frame metadata
camera_frames	Tensor containing projected 3D bbox coordinates
boxes_3d	Tensor containing 3D bounding box coordinates
labels_3d	Tensor containing label indices
instance_ids	Tensor containing instance IDs
scores_3d	Tensor containing confidence scores
img_width	Width of the image
img_height	Height of the image
class_labels	Vector of class label strings
bbox_color_mode	Color mode for bounding boxes (class_id or instance_id)
display_class_labels	Whether to display class labels
display_instance_ids	Whether to display instance IDs
display_class_score	Whether to display class confidence score
logger	Logger for debug and warning messages

calculate_centroid()

void calculate_centroid	(	const float *	points,
		float *	centroid
	)

Calculates the centroid of a 3D bounding box from its 2D projection.

Parameters

points	Array of 2D points
centroid	Array to store centroid coordinates

clamp_point()

void clamp_point	(	float *	x,
		float *	y,
		int	img_width,
		int	img_height
	)

Clamps a point to stay within image boundaries.

Parameters

x	X-coordinate to clamp
y	Y-coordinate to clamp
img_width	Image width
img_height	Image height

count_visible_corners()

bool count_visible_corners	(	const float *	points,
		int	num_corners,
		int	img_width,
		int	img_height
	)

Counts visible corners of a 3D bounding box.

Parameters

points	Array of 2D points
num_corners	Number of corners
img_width	Image width
img_height	Image height

Returns

true if half or more corners are visible

extract_class_labels()

std::vector<std::string> extract_class_labels

(

const std::vector< ClassThreshold > &

class_thresholds

)

Extract class label strings from ClassThreshold vector.

Parameters

class_thresholds

Vector of ClassThreshold structures

Returns

Vector of class label strings only

fetch_class_labels_with_thresholds()

std::vector<ClassThreshold> fetch_class_labels_with_thresholds	(	const std::string &	labels_file,
		PrettyLogger *	logger
	)

Fetch class labels and thresholds from a labels file.

This function reads class labels and their confidence thresholds from a file where each entry is in format: class_name:threshold; Example: Person:0.85;Vehicle:0.90;

Parameters

labels_file	Path to the file containing class labels and thresholds
logger	Logger for debug and warning messages

Returns

Vector of ClassThreshold structures containing labels and thresholds

get_color_from_instance_id()

void get_color_from_instance_id	(	int64_t	instance_id,
		float *	red,
		float *	green,
		float *	blue
	)

Generates a color based on instance ID for tracking visualization.

Parameters

instance_id	Instance ID
red	Output red component [0-1]
green	Output green component [0-1]
blue	Output blue component [0-1]

get_color_from_label()

void get_color_from_label	(	const std::string &	label,
		float *	red,
		float *	green,
		float *	blue,
		int	num_classes
	)

Generates a consistent color based on a class label.

Parameters

label	Class label string
red	Output red component [0-1]
green	Output green component [0-1]
blue	Output blue component [0-1]
num_classes	Number of unique classes