ANSCORE/modules/ANSODEngine/ANSOVFaceDetector.h

#ifndef ANSOVFaceDetector_H
#define ANSOVFaceDetector_H
#pragma once
#include <string.h>
#include <vector>
#include <iostream>
#include <typeinfo>
#include "ANSEngineCommon.h"
#include "cnn.hpp"
#include "ONNXEngine.h"
//#define USEOVFACEDETECTOR
//#define FACEDEBUG 
namespace ANSCENTER {
    class ANSENGINE_API ANSOVFD :public ANSFDBase {
    public:
        virtual bool Initialize(std::string licenseKey, ModelConfig modelConfig, const std::string& modelZipFilePath, const std::string& modelZipPassword, std::string& labelMap) override;
        virtual bool LoadModel(const std::string& modelZipFilePath, const std::string& modelZipPassword)override;
        virtual bool LoadModelFromFolder(std::string licenseKey, ModelConfig modelConfig, std::string modelName, std::string className,const std::string& modelFolder, std::string& labelMap)override;
        virtual bool OptimizeModel(bool fp16, std::string& optimizedModelFolder);
        std::vector<Object> RunInference(const cv::Mat& input, bool useDynamicImage = true, bool validateFace=false, bool facelivenessCheck=true);
        std::vector<Object> RunInference(const cv::Mat& input, const std::string& camera_id, bool useDynamicImage = true, bool validateFace=false, bool facelivenessCheck=true);
        bool Destroy();
        ~ANSOVFD();
    private:
        std::mutex              _mtx;
        std::string             _modelFilePath;
        std::unique_ptr<SCRFD>  _face_detector = nullptr;

        // DML device-lost recovery (see ANSONNXYOLO.h for rationale)
        bool _dmlDeviceLost{ false };
        std::string _scrfdModelPath;  // cached for CPU fallback recreation
        std::vector<Object>     _movementObjects;
		int                     _retainDetectedFaces{ 0 };
        std::vector<Object> Inference(const cv::Mat& input, const std::string& camera_id, bool useDynamicImage = true, bool validateFace=false);
#ifdef USEOVFACEDETECTOR
        std::string         _landmarkModelFilePath;
		std::unique_ptr<VectorCNN> _faceLandmark = nullptr;
        ov::InferRequest request;
        ov::Tensor inTensor;
        ov::Shape inShape;
        ov::CompiledModel compiled_model_;
        const size_t ndetections = 200;
        std::string output;
        std::string labels_output;
        bool _useYoloFaceDetector{true};
        size_t objectSize;
        float width;
        float height;
        size_t model_input_width;
        size_t model_input_height;
        float bb_enlarge_coefficient;
        float bb_dx_coefficient;
        float bb_dy_coefficient;
		std::shared_ptr<ov::Model> read(const ov::Core& core, std::string pathToModel);
        std::vector<Object>  runOVInference(const cv::Mat& frame, const std::string& cameraId);
        void InitialSSDModel(const std::string& deviceName);
        void InitialYoloModel(const std::string& deviceName);
        cv::Mat PreProcessing(const cv::Mat& source);
        std::vector<Object>  runYoloInference(const cv::Mat& input, const std::string& cameraId);
        cv::Mat resized_frame_;
        cv::Point2f factor_;
        cv::Size2f model_input_shape_;
        cv::Size model_output_shape_;
		bool InitFaceDetector();
#endif
    };
}
#endif