WSeries_Calibration/QtCameraHardWareCopilot/include/Vision.h

#pragma once

#ifdef ORTTOOLKIT
#define ORTTOOLKIT __declspec(dllexport)
#else
#define ORTTOOLKIT __declspec(dllimport)
#endif

#include <vector>
#include <array>
#include <opencv/cv.h>
#include <onnxruntime/onnxruntime_cxx_api.h>
#include "Type.h"

namespace OrtToolkit {
	namespace Vision {
		namespace Detect {
			/// <summary>
			/// 飞桨PP_YOLOE转ONNX模型BatchSize只能为1
			/// </summary>
			class ORTTOOLKIT PP_YOLOE
			{
			public:
				/// <summary>
				/// PP_YOLOE模型初始化
				/// </summary>
				/// <param name="model_path">模型路径</param>
				/// <param name="confThreshold">置信度阈值</param>
				/// <param name="ExecutionProvider">ONNX概念下的 执行提供程序 (EP),参数为字符串,默认为CPU,其他：CUDA</param>
				PP_YOLOE(std::string model_path, float confThreshold, std::string ExecutionProvider = "CUDA");
				/// <summary>
				/// 目标检测
				/// </summary>
				/// <param name="cv_image">输入OpenCV的BGR图像</param>
				void Predict(cv::Mat& cv_image, Type::Res_PP_YOLOE& res);
			private:
				float confThreshold;//置信度阈值
				cv::Mat preprocess(cv::Mat srcimg);
				const int inpWidth = 640;
				const int inpHeight = 640;
				std::vector<float> mean_ = { 0.485, 0.456, 0.406 };
				std::vector<float> std_ = { 0.229, 0.224, 0.225 };
				std::vector<float> input_image_;
				std::vector<float> scale_factor = { 1,1 };
				Ort::Env env = Ort::Env(ORT_LOGGING_LEVEL_ERROR, "PP_YOLOE");
				Ort::Session* ort_session = nullptr;
				Ort::SessionOptions sessionOptions = Ort::SessionOptions();
				std::vector<char*> input_names;
				std::vector<char*> output_names;
				std::vector<std::vector<int64_t>> input_node_dims; // 输入节点维度
				std::vector<std::vector<int64_t>> output_node_dims; // 输出节点维度
			};
		}
		namespace FaceDetect {
			class ORTTOOLKIT YOLOV7_face
			{
			public:
				YOLOV7_face(float confThreshold, float nmsThreshold, std::string modelpath, std::string ExecutionProvider = "CUDA");
				void Predict(cv::Mat& frame, Type::Res_YOLOV7_FACE& generate_boxes);
			private:
				int inpWidth;
				int inpHeight;
				int nout;
				int num_proposal;
				float confThreshold;//置信度阈值
				float nmsThreshold;//nms阈值
				std::vector<float> mean_ = { 0.5,0.5,0.5 };
				std::vector<float> std_ = { 0.5,0.5,0.5 };
				std::vector<float> input_image_;//模型输入图片
				void nms(std::vector<Type::Type_Iou_YOLOV7FACE>Items, Type::Res_YOLOV7_FACE& Res);//NMS
				bool has_postprocess;
				std::vector<Ort::AllocatedStringPtr>In_AllocatedStringPtr;
				std::vector<Ort::AllocatedStringPtr>Out_AllocatedStringPtr;
				Ort::Env env = Ort::Env(ORT_LOGGING_LEVEL_ERROR, "YOLOV7_face");
				Ort::Session* ort_session = nullptr;//OnnxRunTime会话
				Ort::SessionOptions sessionOptions = Ort::SessionOptions();
				std::vector<char*> input_names;
				std::vector<char*> output_names;
				std::vector<std::vector<int64_t>> input_node_dims; // >=1 outputs
				std::vector<std::vector<int64_t>> output_node_dims; // >=1 outputs
			};
		}
		namespace FaceAlign {
			class ORTTOOLKIT PFLD98
			{
			public:
				PFLD98(std::string ModPath);
				void Predict(cv::Mat& Img, std::vector<cv::Point2f>& Density);
				~PFLD98();

			private:
				Ort::Env env = Ort::Env(ORT_LOGGING_LEVEL_WARNING, "MCNN");
				Ort::SessionOptions sessionOptions = Ort::SessionOptions();
				Ort::AllocatorWithDefaultOptions allocator;

				const char* input_name;
				const char* output_nameA;
				const char* output_nameB;

				std::vector<int64_t> input_dims;
				std::vector<int64_t> output_dims;

				Ort::Session* ort_session;
				std::vector<const char*> input_names;
				std::vector<const char*> output_names;
				std::vector<const char*> input_node_names;
				std::vector<const char*> output_node_names;
			private:
				int DensityRow;
				int DensityCol;
				std::vector<float> input_image_;
			};
		}
		namespace FaceId {
			class ORTTOOLKIT Arcface {
			public:
				Arcface(std::string modelpath, std::string ExecutionProvider = "CUDA");
				void Predict(cv::Mat& frame, std::vector<float>& FaceVector);
			private:
				int inpWidth;
				int inpHeight;
				int nout;
				std::vector<float> input_image_;//模型输入图片
				std::vector<float> mean_ = { 0.5, 0.5, 0.5 };
				std::vector<float> std_ = { 0.5, 0.5, 0.5 };
				std::vector<Ort::AllocatedStringPtr>In_AllocatedStringPtr;
				std::vector<Ort::AllocatedStringPtr>Out_AllocatedStringPtr;
				Ort::Env env = Ort::Env(ORT_LOGGING_LEVEL_ERROR, "Arcface");//日志记录
				Ort::Session* ort_session = nullptr;//OnnxRunTime会话
				Ort::SessionOptions sessionOptions = Ort::SessionOptions();
				std::vector<char*> input_names;
				std::vector<char*> output_names;
				std::vector<std::vector<int64_t>> input_node_dims; // >=1 outputs
				std::vector<std::vector<int64_t>> output_node_dims; // >=1 outputs
			};
		}
		namespace CrowdCounting {
			class ORTTOOLKIT MCNN
			{
			public:
				MCNN(std::string ModPath);
				void Predict(cv::Mat& Img, std::vector<float>& Density);
				~MCNN();

			private:
				Ort::Env env = Ort::Env(ORT_LOGGING_LEVEL_WARNING, "MCNN");
				Ort::SessionOptions sessionOptions = Ort::SessionOptions();
				Ort::AllocatorWithDefaultOptions allocator;

				const char* input_name;
				const char* output_name;

				std::vector<int64_t> input_dims;
				std::vector<int64_t> output_dims;

				Ort::Session* ort_session;
				std::vector<const char*> input_names;
				std::vector<const char*> output_names;
				std::vector<const char*> input_node_names;
				std::vector<const char*> output_node_names;
				std::vector<float> input_image_;
			private:
				int DensityRow;
				int DensityCol;
			};
		}
	}
	namespace  Util{ 
		void ORTTOOLKIT overflow(float& info, int lower, int upper);
		void ORTTOOLKIT BoxOverflow(Type::Base_Type_Box& Box, int lower, int upper);
		float ORTTOOLKIT CosineSimilarity(const std::vector<float>& a, const std::vector<float>& b, bool normalized = false);
		std::string ORTTOOLKIT create_uuid();
		uint16_t floatToHalf(float f);
		float halfToFloat(uint16_t h);
		template <typename T, typename U>
		void ReadBin(U& x, std::string path) {
			std::ifstream file(path, std::ios::binary);
			if (file) {
				file.seekg(0, std::ios::end);
				std::streampos size = file.tellg();
				file.seekg(0, std::ios::beg);
				x.resize(size / sizeof(T));
				file.read(reinterpret_cast<char*>(x.data()), size);
				file.close();
			}
		};
	}
}