yolomodel Namespace Reference

Namespace containing functions or objects/struct used to aid in easy use of YOLO models. Namespace can contain other namespace pertaining to the method of inference or library being used. More...

Namespaces
namespace	pytorch
	Namespace containing functions or objects/structs used to run inference on a YOLO model with the PyTorch library. This namespace contains static functions for getting available hardware devices, and classes for running a .pt model on each device. This namespace was built to work with YOLO models only!
namespace	tensorflow
	Namespace containing functions or objects/structs used to run inference on a YOLO model with the Tensorflow library. This namespace contains static functions for getting available hardware devices, and classes for running a .tflite model on each device. This namespace was built to work with YOLO models only!

Classes
struct	Detection
	This struct is used to. More...

Functions
void	NonMaxSuppression (std::vector< Detection > &vObjects, std::vector< int > &vClassIDs, std::vector< float > &vClassConfidences, std::vector< cv::Rect > &vBoundingBoxes, float fMinObjectConfidence, float fNMSThreshold)
	Perform non max suppression for the given predictions. This eliminates/combines predictions that overlap with each other.
void	DrawDetections (cv::Mat &cvInputFrame, std::vector< Detection > &vObjects)
	Given an image and a vector of object structs, draw each object bounding box, class type, and confidence onto the image.

Detailed Description

Namespace containing functions or objects/struct used to aid in easy use of YOLO models. Namespace can contain other namespace pertaining to the method of inference or library being used.

Author

clayjay3 (clayt.nosp@m.onra.nosp@m.ycowe.nosp@m.n@gm.nosp@m.ail.c.nosp@m.om)

Date

2023-11-11

Function Documentation

NonMaxSuppression()

void yolomodel::NonMaxSuppression ( std::vector< Detection > &  vObjects, std::vector< int > &  vClassIDs, std::vector< float > &  vClassConfidences, std::vector< cv::Rect > &  vBoundingBoxes, float  fMinObjectConfidence, float  fNMSThreshold  )

inline

Perform non max suppression for the given predictions. This eliminates/combines predictions that overlap with each other.

Parameters

vObjects	- A vector that will be filled with all of the valid/filtered predictions with their data stored in an easy-to-use struct.
vClassIDs	- A reference to a vector that contains class IDs for each prediction.
vClassConfidences	- A reference to a vector that contains the highest class confidence for that prediction.
vBoundingBoxes	- A reference to a vector that contains a cv::Rect bounding box for each prediction.
fMinObjectConfidence	- The minimum confidence for determining which predictions to throw out.
fNMSThreshold	- The threshold value for filtering out weaker bounding boxes or detections.

Author

clayjay3 (clayt.nosp@m.onra.nosp@m.ycowe.nosp@m.n@gm.nosp@m.ail.c.nosp@m.om)

Date

2023-11-15

    {
        // Create instance variables.
        std::vector<int> vNMSValidIndices;
 
        // Perform Non-Max Suppression using OpenCV's implementation.
        cv::dnn::NMSBoxes(vBoundingBoxes, vClassConfidences, fMinObjectConfidence, fNMSThreshold, vNMSValidIndices);
 
        // Loop through each valid index.
        for (int nValidIndex : vNMSValidIndices)
        {
            // Create new Detection struct.
            Detection stNewDetection;
            // Repackage prediction data into easy-to-use struct.
            stNewDetection.nClassID      = vClassIDs[nValidIndex];
            stNewDetection.fConfidence   = vClassConfidences[nValidIndex];
            stNewDetection.cvBoundingBox = vBoundingBoxes[nValidIndex];
 
            // Append new object detection to objects vector.
            vObjects.emplace_back(stNewDetection);
        }
    }

Here is the call graph for this function:

Here is the caller graph for this function:

DrawDetections()

void yolomodel::DrawDetections ( cv::Mat &  cvInputFrame, std::vector< Detection > &  vObjects  )

inline

Given an image and a vector of object structs, draw each object bounding box, class type, and confidence onto the image.

Parameters

cvInputFrame	- A reference to the cv::Mat to draw overlays on.
vObjects	- A reference to the vector containing the object detection structs.

Author

clayjay3 (clayt.nosp@m.onra.nosp@m.ycowe.nosp@m.n@gm.nosp@m.ail.c.nosp@m.om)

Date

2023-11-15

    {
        // Loop through each detection.
        for (Detection stObject : vObjects)
        {
            // Calculate the hue value based on the class ID.
            int nHue = static_cast<int>(stObject.nClassID % 256);
            // Set saturation and value to 1.0 for full intensity colors.
            int nSaturation = 255;
            int nValue      = 255;
 
            // Convert HSV to RGB
            cv::Mat cvHSV(1, 1, CV_8UC3, cv::Scalar(nHue, nSaturation, nValue));
            cv::cvtColor(cvHSV, cvHSV, cv::COLOR_HSV2BGR);
            // Extract the RGB values
            cv::Vec3b cvConvertedValues = cvHSV.at<cv::Vec3b>(0, 0);
            cv::Scalar cvBoxColor(cvConvertedValues[2], cvConvertedValues[1], cvConvertedValues[0]);
 
            // Draw bounding box onto image.
            cv::rectangle(cvInputFrame, stObject.cvBoundingBox, cvBoxColor, 2);
            // Draw classID background box onto image.
            cv::rectangle(cvInputFrame,
                          cv::Point(stObject.cvBoundingBox.x, stObject.cvBoundingBox.y - 20),
                          cv::Point(stObject.cvBoundingBox.x + stObject.cvBoundingBox.width, stObject.cvBoundingBox.y),
                          cvBoxColor,
                          cv::FILLED);
            // Draw class text onto image.
            cv::putText(cvInputFrame,
                        std::to_string(stObject.nClassID) + " " + std::to_string(stObject.fConfidence),
                        cv::Point(stObject.cvBoundingBox.x, stObject.cvBoundingBox.y - 5),
                        cv::FONT_HERSHEY_SIMPLEX,
                        0.5,
                        cv::Scalar(255, 255, 255));
        }
    }

Here is the call graph for this function:

Here is the caller graph for this function: