170 lines
8.8 KiB
C
170 lines
8.8 KiB
C
|
#ifndef ANSOCRBASE_H
|
||
|
|
#define ANSOCRBASE_H
|
||
|
|
#define ANSOCR_API __declspec(dllexport)
|
||
|
|
#pragma once
|
||
|
|
#include "opencv2/core.hpp"
|
||
|
|
#include "opencv2/imgcodecs.hpp"
|
||
|
|
#include "opencv2/imgproc.hpp"
|
||
|
|
#include <iostream>
|
||
|
|
#include <vector>
|
||
|
|
#include "LabVIEWHeader/extcode.h"
|
||
|
|
#include "ANSLicense.h"
|
||
|
|
namespace ANSCENTER {
|
||
|
|
struct OCRModelConfig {
|
||
|
|
bool userGPU = true;
|
||
|
|
bool useTensorRT = false;
|
||
|
|
int gpuId = 0;
|
||
|
|
int gpuMemory = 4000;
|
||
|
|
int cpuThreads = 10;
|
||
|
|
bool enableMKLDNN = false;
|
||
|
|
bool ensureASCII = true;
|
||
|
|
OCRLanguage ocrLanguage;
|
||
|
|
std::string precisionType;
|
||
|
|
std::string ocrType;
|
||
|
|
std::string limitType;
|
||
|
|
|
||
|
|
bool useDetector = true;
|
||
|
|
bool useRecognizer = true;
|
||
|
|
bool useCLS = true;
|
||
|
|
bool useTable = false;
|
||
|
|
bool useLayout = false;
|
||
|
|
|
||
|
|
std::string detectionModelDir;
|
||
|
|
std::string detectionModelFile;
|
||
|
|
std::string detectionModelParam;
|
||
|
|
std::string detectionScoreMode;
|
||
|
|
|
||
|
|
|
||
|
|
std::string recognizerModelDir;
|
||
|
|
std::string recognizerModelFile;
|
||
|
|
std::string recognizerModelParam;
|
||
|
|
std::string recogizerCharDictionaryPath;
|
||
|
|
|
||
|
|
std::string clsModelDir;
|
||
|
|
std::string clsModelFile;
|
||
|
|
std::string clsModelParam;
|
||
|
|
|
||
|
|
|
||
|
|
std::string layoutModelDir;
|
||
|
|
std::string layourDictionaryPath;
|
||
|
|
std::string tableModelDir;
|
||
|
|
std::string tableCharDictionaryPath;
|
||
|
|
|
||
|
|
int limitSideLen = 960;
|
||
|
|
double detectionDBThreshold = 0.3;
|
||
|
|
double detectionBoxThreshold = 0.6;
|
||
|
|
double detectionDBUnclipRatio = 1.5;
|
||
|
|
bool useDilation = false;
|
||
|
|
bool useAngleCLS = false;
|
||
|
|
double clsThreshold = 0.9;
|
||
|
|
int clsBatchNumber = 1;
|
||
|
|
int recognizerBatchNum = 6;
|
||
|
|
int recoginzerImageHeight = 48;
|
||
|
|
int recoginzerImageWidth = 320;
|
||
|
|
double layoutScoreThreshold = 0.5;
|
||
|
|
double layoutNMSThreshold = 0.5;
|
||
|
|
int tableModelMaxLengh = 488;
|
||
|
|
int tableBatchNum = 1;
|
||
|
|
bool mergeNoSpanStructure = true;
|
||
|
|
};
|
||
|
|
struct OCRObject
|
||
|
|
{
|
||
|
|
int classId{ 0 };
|
||
|
|
int trackId{ 0 };
|
||
|
|
std::string className{};
|
||
|
|
float confidence{ 0.0 };
|
||
|
|
cv::Rect box{};
|
||
|
|
std::vector<cv::Point2f> polygon; // polygon that contain x1 ,y1,x2,y2,x3,y3,x4,y4
|
||
|
|
std::vector<float> kps{}; // Containing keypoints
|
||
|
|
cv::Mat mask{}; // jpeg string of the mask
|
||
|
|
std::string extraInfo; // More information such as facial recognition
|
||
|
|
std::string cameraId;
|
||
|
|
//std::string attributes; // Attributes information in JSON string
|
||
|
|
};
|
||
|
|
class ANSOCR_API ANSOCRBase {
|
||
|
|
protected:
|
||
|
|
bool _licenseValid{ false };
|
||
|
|
bool _isInitialized{ false };
|
||
|
|
std::string _licenseKey;
|
||
|
|
std::string _modelFolder;
|
||
|
|
std::string _modelConfigFile;
|
||
|
|
OCRModelConfig _modelConfig;
|
||
|
|
int _engineMode; //0: Auto detect, 1 GPU, 2 CPU
|
||
|
|
SPDLogger& _logger = SPDLogger::GetInstance("OCR", false);
|
||
|
|
void CheckLicense();
|
||
|
|
[[nodiscard]] bool Init(const std::string& licenseKey, OCRModelConfig modelConfig, const std::string& modelZipFilePath, const std::string& modelZipPassword, int engineMode);
|
||
|
|
public:
|
||
|
|
[[nodiscard]] virtual bool Initialize(const std::string& licenseKey, OCRModelConfig modelConfig, const std::string& modelZipFilePath, const std::string& modelZipPassword, int engineMode);
|
||
|
|
[[nodiscard]] virtual std::vector<ANSCENTER::OCRObject> RunInference(const cv::Mat& input) = 0;
|
||
|
|
[[nodiscard]] virtual std::vector<ANSCENTER::OCRObject> RunInference(const cv::Mat& input, const std::string& cameraId) = 0;
|
||
|
|
[[nodiscard]] virtual std::vector<ANSCENTER::OCRObject> RunInference(const cv::Mat& input, const std::vector<cv::Rect>& Bbox) = 0;
|
||
|
|
[[nodiscard]] virtual std::vector<ANSCENTER::OCRObject> RunInference(const cv::Mat& input, const std::vector<cv::Rect>& Bbox, const std::string &cameraId) = 0;
|
||
|
|
|
||
|
|
~ANSOCRBase() {
|
||
|
|
try {
|
||
|
|
|
||
|
|
}
|
||
|
|
catch (std::exception& e) {
|
||
|
|
std::cout << "ANSOCRBase::Destroy()" << e.what();
|
||
|
|
}
|
||
|
|
|
||
|
|
};
|
||
|
|
[[nodiscard]] virtual bool Destroy() = 0;
|
||
|
|
};
|
||
|
|
class ANSOCRUtility
|
||
|
|
{
|
||
|
|
public:
|
||
|
|
[[nodiscard]] static std::string OCRDetectionToJsonString(const std::vector<OCRObject>& dets);
|
||
|
|
[[nodiscard]] static std::vector<cv::Rect> GetBoundingBoxes(const std::string& strBBoxes);
|
||
|
|
[[nodiscard]] static std::string PolygonToString(const std::vector<cv::Point2f>& polygon);
|
||
|
|
[[nodiscard]] static std::vector<cv::Point2f> RectToNormalizedPolygon(const cv::Rect& rect, float imageWidth, float imageHeight);
|
||
|
|
[[nodiscard]] static std::string KeypointsToString(const std::vector<float>& kps);
|
||
|
|
private:
|
||
|
|
};
|
||
|
|
}
|
||
|
|
|
||
|
|
// Original signature — backward compatible with third-party apps built against older DLL
|
||
|
|
extern "C" ANSOCR_API int CreateANSOCRHandle(ANSCENTER::ANSOCRBase** Handle, const char* licenseKey, const char* modelFilePath,
|
||
|
|
const char* modelFileZipPassword, int language, int engineMode, int gpuId = 0,
|
||
|
|
double detectorDBThreshold = 0.3, double detectorDBBoxThreshold = 0.6, double detectorDBUnclipRatio = 1.5,
|
||
|
|
double classifierThreshold = 0.9, int useDilation = 0);
|
||
|
|
// Extended version with limitSideLen parameter — new callers should use this
|
||
|
|
extern "C" ANSOCR_API int CreateANSOCRHandleEx(ANSCENTER::ANSOCRBase** Handle, const char* licenseKey, const char* modelFilePath,
|
||
|
|
const char* modelFileZipPassword, int language, int engineMode, int gpuId = 0,
|
||
|
|
double detectorDBThreshold = 0.3, double detectorDBBoxThreshold = 0.6, double detectorDBUnclipRatio = 1.5,
|
||
|
|
double classifierThreshold = 0.9, int useDilation = 0, int limitSideLen = 960);
|
||
|
|
extern "C" ANSOCR_API std::string RunInference(ANSCENTER::ANSOCRBase** Handle, unsigned char* jpeg_string, int32 bufferLength);
|
||
|
|
extern "C" ANSOCR_API std::string RunInferenceWithCamID(ANSCENTER::ANSOCRBase** Handle, unsigned char* jpeg_string, int32 bufferLength, const char* cameraId);
|
||
|
|
extern "C" ANSOCR_API int RunInferenceCV(ANSCENTER::ANSOCRBase** Handle, const cv::Mat &image, std::string &ocrResult);
|
||
|
|
|
||
|
|
extern "C" ANSOCR_API std::string RunInferenceBinary(ANSCENTER::ANSOCRBase** Handle, unsigned char* jpeg_bytes, unsigned int width, unsigned int height);
|
||
|
|
extern "C" ANSOCR_API int ReleaseANSOCRHandle(ANSCENTER::ANSOCRBase** Handle);
|
||
|
|
extern "C" ANSOCR_API std::string RunInferenceImagePath(ANSCENTER::ANSOCRBase** Handle, const char* imageFilePath);
|
||
|
|
extern "C" ANSOCR_API std::string RunInferenceInCroppedImages(ANSCENTER::ANSOCRBase** Handle, unsigned char* jpeg_string, int32 bufferLength, const char* strBboxes);
|
||
|
|
extern "C" ANSOCR_API std::string RunInferenceInCroppedImagesWithCamID(ANSCENTER::ANSOCRBase** Handle, unsigned char* jpeg_string, int32 bufferLength, const char* strBboxes, const char* cameraId);
|
||
|
|
|
||
|
|
//// For LabVIEW API
|
||
|
|
extern "C" ANSOCR_API int RunInference_LV(ANSCENTER::ANSOCRBase** Handle, unsigned char* jpeg_string, int32 bufferLength, LStrHandle detectionResult);
|
||
|
|
extern "C" ANSOCR_API int RunInference_LVWithCamID(ANSCENTER::ANSOCRBase** Handle, unsigned char* jpeg_string, int32 bufferLength, const char* cameraId,LStrHandle detectionResult);
|
||
|
|
|
||
|
|
extern "C" ANSOCR_API int RunInferenceBinary_LV(ANSCENTER::ANSOCRBase** Handle, unsigned char* jpeg_bytes, unsigned int width, unsigned int height, LStrHandle detectionResult);
|
||
|
|
extern "C" ANSOCR_API int RunInferenceImagePath_LV(ANSCENTER::ANSOCRBase** Handle, const char* imageFilePath, LStrHandle detectionResult);
|
||
|
|
extern "C" ANSOCR_API int ANSOCRUnitTest(const char* modelFilePath, const char* imageFilePath, LStrHandle detectionResult);
|
||
|
|
extern "C" ANSOCR_API int RunInferenceInCroppedImages_LV(ANSCENTER::ANSOCRBase** Handle, unsigned char* jpeg_string, int32 bufferLength, const char* strBboxes, LStrHandle detectionResult);
|
||
|
|
extern "C" ANSOCR_API int RunInferenceInCroppedImages_LVWithCamID(ANSCENTER::ANSOCRBase** Handle, unsigned char* jpeg_string, int32 bufferLength, const char* strBboxes, const char* cameraId, LStrHandle detectionResult);
|
||
|
|
extern "C" ANSOCR_API int RunInferenceComplete_LV(ANSCENTER::ANSOCRBase** Handle, cv::Mat** cvImage, const char* cameraId, int getJpegString, int jpegImageSize, LStrHandle detectionResult, LStrHandle imageStr);
|
||
|
|
extern "C" ANSOCR_API int RunInferencesComplete_LV(ANSCENTER::ANSOCRBase** Handle, cv::Mat** cvImage, const char* cameraId, int maxImageSize, const char* strBboxes, LStrHandle detectionResult);
|
||
|
|
|
||
|
|
// V2 Create / Release — handle as uint64_t by value (no pointer-to-pointer)
|
||
|
|
extern "C" ANSOCR_API uint64_t CreateANSOCRHandleEx_V2(const char* licenseKey, const char* modelFilePath,
|
||
|
|
const char* modelFileZipPassword, int language, int engineMode, int gpuId,
|
||
|
|
double detectorDBThreshold, double detectorDBBoxThreshold, double detectorDBUnclipRatio,
|
||
|
|
double classifierThreshold, int useDilation, int limitSideLen);
|
||
|
|
extern "C" ANSOCR_API uint64_t CreateANSOCRHandle_V2(const char* licenseKey, const char* modelFilePath,
|
||
|
|
const char* modelFileZipPassword, int language, int engineMode, int gpuId,
|
||
|
|
double detectorDBThreshold, double detectorDBBoxThreshold, double detectorDBUnclipRatio,
|
||
|
|
double classifierThreshold, int useDilation);
|
||
|
|
extern "C" ANSOCR_API int ReleaseANSOCRHandle_V2(uint64_t handleVal);
|
||
|
|
|
||
|
|
#endif
|