Collaboration diagram for apollo::perception::lidar::PostprocessCuda:

Public Member Functions
	PostprocessCuda (const float float_min, const float float_max, const int num_anchor, const int num_class, const float score_threshold, const int num_threads, const float nms_overlap_threshold, const int num_box_corners, const int num_output_box_feature)
	Constructor. More...

void	DoPostprocessCuda (const float rpn_box_output, const float rpn_cls_output, const float rpn_dir_output, int dev_anchor_mask, const float dev_anchors_px, const float dev_anchors_py, const float dev_anchors_pz, const float dev_anchors_dx, const float dev_anchors_dy, const float dev_anchors_dz, const float dev_anchors_ro, float dev_filtered_box, float dev_filtered_score, int dev_filtered_label, int dev_filtered_dir, float dev_box_for_nms, int dev_filter_count, std::vector< float > out_detection, std::vector< int > *out_label)
	Postprocessing for the network output. More...

Constructor & Destructor Documentation

◆ PostprocessCuda()

apollo::perception::lidar::PostprocessCuda::PostprocessCuda	(	const float	float_min,
		const float	float_max,
		const int	num_anchor,
		const int	num_class,
		const float	score_threshold,
		const int	num_threads,
		const float	nms_overlap_threshold,
		const int	num_box_corners,
		const int	num_output_box_feature
	)

Constructor.

Parameters

[in]	float_min	The lowest float value
[in]	float_max	The maximum float value
[in]	num_anchor	Number of anchors in total
[in]	num_class	Number of object's classes
[in]	score_threshold	Score threshold for filtering output
[in]	num_threads	Number of threads when launching cuda kernel
[in]	nms_overlap_threshold	IOU threshold for NMS
[in]	num_box_corners	Number of box's corner
[in]	num_output_box_feature	Number of output box's feature

Captital variables never change after the compile, non-capital variables could be changed through rosparam

void apollo::perception::lidar::PostprocessCuda::DoPostprocessCuda	(	const float *	rpn_box_output,
		const float *	rpn_cls_output,
		const float *	rpn_dir_output,
		int *	dev_anchor_mask,
		const float *	dev_anchors_px,
		const float *	dev_anchors_py,
		const float *	dev_anchors_pz,
		const float *	dev_anchors_dx,
		const float *	dev_anchors_dy,
		const float *	dev_anchors_dz,
		const float *	dev_anchors_ro,
		float *	dev_filtered_box,
		float *	dev_filtered_score,
		int *	dev_filtered_label,
		int *	dev_filtered_dir,
		float *	dev_box_for_nms,
		int *	dev_filter_count,
		std::vector< float > *	out_detection,
		std::vector< int > *	out_label
	)

Postprocessing for the network output.

Parameters

[in]	rpn_box_output	Box predictions from the network output
[in]	rpn_cls_output	Class predictions from the network output
[in]	rpn_dir_output	Direction predictions from the network output
[in]	dev_anchor_mask	Anchor mask for filtering the network output
[in]	dev_anchors_px	X-coordinate values for corresponding anchors
[in]	dev_anchors_py	Y-coordinate values for corresponding anchors
[in]	dev_anchors_pz	Z-coordinate values for corresponding anchors
[in]	dev_anchors_dx	X-dimension values for corresponding anchors
[in]	dev_anchors_dy	Y-dimension values for corresponding anchors
[in]	dev_anchors_dz	Z-dimension values for corresponding anchors
[in]	dev_anchors_ro	Rotation values for corresponding anchors
[in]	dev_filtered_box	Filtered box predictions
[in]	dev_filtered_score	Filtered score predictions
[in]	dev_filtered_label	Filtered label predictions
[in]	dev_filtered_dir	Filtered direction predictions
[in]	dev_box_for_nms	Decoded boxes in min_x min_y max_x max_y represenation from pose and dimension
[in]	dev_filter_count	The number of filtered output
[out]	out_detection	Output bounding boxes
[out]	out_label	Output labels of objects

dev_* represents device memory allocated variables

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

modules/perception/lidar/lib/detector/point_pillars_detection/postprocess_cuda.h