TensorRT 8.5.1
nvinfer1::plugin::DetectionOutputParameters Struct Reference

The DetectionOutput plugin layer generates the detection output based on location and confidence predictions by doing non maximum suppression. This plugin first decodes the bounding boxes based on the anchors generated. It then performs non_max_suppression on the decoded bounding boxes. DetectionOutputParameters defines a set of parameters for creating the DetectionOutput plugin layer. It contains: More...

#include <NvInferPluginUtils.h>

Public Attributes

bool shareLocation
 
bool varianceEncodedInTarget
 
int32_t backgroundLabelId
 
int32_t numClasses
 
int32_t topK
 
int32_t keepTopK
 
float confidenceThreshold
 
float nmsThreshold
 
CodeTypeSSD codeType
 
int32_t inputOrder [3]
 
bool confSigmoid
 
bool isNormalized
 
bool isBatchAgnostic {true}
 

Detailed Description

The DetectionOutput plugin layer generates the detection output based on location and confidence predictions by doing non maximum suppression. This plugin first decodes the bounding boxes based on the anchors generated. It then performs non_max_suppression on the decoded bounding boxes. DetectionOutputParameters defines a set of parameters for creating the DetectionOutput plugin layer. It contains:

Parameters
shareLocationIf true, bounding box are shared among different classes.
varianceEncodedInTargetIf true, variance is encoded in target. Otherwise we need to adjust the predicted offset accordingly.
backgroundLabelIdBackground label ID. If there is no background class, set it as -1.
numClassesNumber of classes to be predicted.
topKNumber of boxes per image with top confidence scores that are fed into the NMS algorithm.
keepTopKNumber of total bounding boxes to be kept per image after NMS step.
confidenceThresholdOnly consider detections whose confidences are larger than a threshold.
nmsThresholdThreshold to be used in NMS.
codeTypeType of coding method for bbox.
inputOrderSpecifies the order of inputs {loc_data, conf_data, priorbox_data}.
confSigmoidSet to true to calculate sigmoid of confidence scores.
isNormalizedSet to true if bounding box data is normalized by the network.
isBatchAgnosticDefaults to true. Set to false if prior boxes are unique per batch

Member Data Documentation

◆ backgroundLabelId

int32_t nvinfer1::plugin::DetectionOutputParameters::backgroundLabelId

◆ codeType

CodeTypeSSD nvinfer1::plugin::DetectionOutputParameters::codeType

◆ confidenceThreshold

float nvinfer1::plugin::DetectionOutputParameters::confidenceThreshold

◆ confSigmoid

bool nvinfer1::plugin::DetectionOutputParameters::confSigmoid

◆ inputOrder

int32_t nvinfer1::plugin::DetectionOutputParameters::inputOrder[3]

◆ isBatchAgnostic

bool nvinfer1::plugin::DetectionOutputParameters::isBatchAgnostic {true}

◆ isNormalized

bool nvinfer1::plugin::DetectionOutputParameters::isNormalized

◆ keepTopK

int32_t nvinfer1::plugin::DetectionOutputParameters::keepTopK

◆ nmsThreshold

float nvinfer1::plugin::DetectionOutputParameters::nmsThreshold

◆ numClasses

int32_t nvinfer1::plugin::DetectionOutputParameters::numClasses

◆ shareLocation

bool nvinfer1::plugin::DetectionOutputParameters::shareLocation

◆ topK

int32_t nvinfer1::plugin::DetectionOutputParameters::topK

◆ varianceEncodedInTarget

bool nvinfer1::plugin::DetectionOutputParameters::varianceEncodedInTarget

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