You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
136 lines
4.1 KiB
136 lines
4.1 KiB
#ifndef __SMARTVISION_EXAMPLE__
|
|
#define __SMARTVISION_EXAMPLE__
|
|
|
|
#include <iostream>
|
|
#include <string>
|
|
#include <opencv2/opencv.hpp>
|
|
#include <aidlite.hpp>
|
|
#include "AIObjectDet.hpp"
|
|
|
|
enum AI_Model_Type_
|
|
{
|
|
AI_Model_Type_Ocr,
|
|
AI_Model_Type_Object,
|
|
AI_Model_Type_Align,
|
|
AI_Model_Type_Seg,
|
|
AI_Model_Type_Reconstruct,
|
|
AI_Model_Type_EdgeAngle,
|
|
AI_Model_Type_Count,
|
|
};
|
|
enum AI_MODEL_TYPE_
|
|
{
|
|
AI_MODEL_TYPE_CPU,
|
|
AI_MODEL_TYPE_INT8,
|
|
AI_MODEL_TYPE_FLOAT32,
|
|
AI_MODEL_TYPE_FLOAT16,
|
|
};
|
|
enum AI_DATA_IDX_
|
|
{
|
|
AI_DATA_IDX_T,
|
|
AI_DATA_IDX_HEIGHT,
|
|
AI_DATA_IDX_WIDTH,
|
|
AI_DATA_IDX_CHANNELS,
|
|
AI_DATA_IDX_COUNT,
|
|
};
|
|
struct AI_ConfigInfo
|
|
{
|
|
std::string strdetect_model_path; // 模型路径
|
|
int ai_Model_Type; // 模型使用类型 AI_MODEL_TYPE_
|
|
bool useMean_Div; // 是否要做均值方差处理
|
|
std::string strName_Out; // 模型输出名称
|
|
unsigned int dataArr_In[AI_DATA_IDX_COUNT]; // 输入 数据尺寸大小
|
|
unsigned int dataArr_Out[AI_DATA_IDX_COUNT]; // 输出 数据尺寸大小
|
|
unsigned int dataArr_Out1[AI_DATA_IDX_COUNT]; // 输出 数据尺寸大小
|
|
cv::Scalar means_scale; // 均值
|
|
cv::Scalar stds_scale; // 方差
|
|
int outImgChannlse; // 输出图片的通道数，如果是单通道，就 只取第一通道的图片，如果是多通道，就全取所有输出通道
|
|
float fmaskThreshold; // mask 输出的阈值
|
|
int userflag;
|
|
AI_ConfigInfo()
|
|
{
|
|
strdetect_model_path = "";
|
|
ai_Model_Type = AI_MODEL_TYPE_FLOAT16;
|
|
useMean_Div = true;
|
|
strName_Out = "";
|
|
dataArr_In[AI_DATA_IDX_T] = 1;
|
|
dataArr_In[AI_DATA_IDX_HEIGHT] = 480;
|
|
dataArr_In[AI_DATA_IDX_WIDTH] = 480;
|
|
dataArr_In[AI_DATA_IDX_COUNT] = 3;
|
|
|
|
dataArr_Out[AI_DATA_IDX_T] = 1;
|
|
dataArr_Out[AI_DATA_IDX_HEIGHT] = 480;
|
|
dataArr_Out[AI_DATA_IDX_WIDTH] = 480;
|
|
dataArr_Out[AI_DATA_IDX_COUNT] = 4;
|
|
|
|
means_scale = cv::Scalar(0.485, 0.456, 0.406);
|
|
stds_scale = cv::Scalar(0.229, 0.224, 0.225);
|
|
outImgChannlse = 1;
|
|
fmaskThreshold = 0.5f;
|
|
userflag = 0;
|
|
}
|
|
void copy(AI_ConfigInfo tem)
|
|
{
|
|
this->strdetect_model_path = tem.strdetect_model_path;
|
|
this->ai_Model_Type = tem.ai_Model_Type;
|
|
this->useMean_Div = tem.useMean_Div;
|
|
this->strName_Out = tem.strName_Out;
|
|
for (int i = 0; i < AI_DATA_IDX_COUNT; i++)
|
|
{
|
|
this->dataArr_In[i] = tem.dataArr_In[i];
|
|
this->dataArr_Out[i] = tem.dataArr_Out[i];
|
|
}
|
|
|
|
this->means_scale = tem.means_scale;
|
|
this->stds_scale = tem.stds_scale;
|
|
this->outImgChannlse = tem.outImgChannlse;
|
|
this->fmaskThreshold = tem.fmaskThreshold;
|
|
this->userflag = tem.userflag;
|
|
}
|
|
};
|
|
|
|
struct ObjectInfo
|
|
{
|
|
cv::Rect roi;
|
|
int classIdx;
|
|
float fscore;
|
|
};
|
|
|
|
class DetectModel
|
|
{
|
|
|
|
public:
|
|
DetectModel();
|
|
~DetectModel();
|
|
|
|
int init(AI_ConfigInfo AIconfig);
|
|
int init_reconstruct(AI_ConfigInfo AIconfig);
|
|
int init_Object(AI_ConfigInfo AIconfig);
|
|
void release();
|
|
int run(const cv::Mat &frame, cv::Mat &maskimg, bool bmeandv);
|
|
int run_Arr_One_Out(const cv::Mat &frame, cv::Mat &maskimg, int imgidx, bool bmeandv, float fT = 0.5);
|
|
int run_arrOut(const cv::Mat &frame, std::vector<cv::Mat> &mask, bool bmeandv, float fT = 0.5);
|
|
int run_Object(const cv::Mat &frame, std::vector<ObjectInfo> &outList, bool bmeandv, float fT = 0.5);
|
|
int run_reconstruct(const cv::Mat &frame, cv::Mat & reconstructimg, bool bmeandv);
|
|
|
|
private:
|
|
int Img_Mean_Div(cv::Mat srcimg, cv::Mat &outimg);
|
|
int UpdataObjectParam(int size);
|
|
|
|
private:
|
|
std::unique_ptr<Aidlux::Aidlite::Interpreter> detect_interpreter;
|
|
|
|
uint8_t *m_DetectOutput;
|
|
|
|
AI_ConfigInfo m_AIconfig;
|
|
|
|
AIObjectDet m_Objectdet;
|
|
// Object
|
|
private:
|
|
int MODEL_SIZE;
|
|
int OBJ_NUMB_MAX_SIZE;
|
|
uint32_t STRIDE8_SIZE;
|
|
uint32_t STRIDE16_SIZE;
|
|
uint32_t STRIDE32_SIZE;
|
|
};
|
|
|
|
#endif |