41 lines
1.1 KiB
C++
41 lines
1.1 KiB
C++
#pragma once
|
|
|
|
#include "ONNXOCRTypes.h"
|
|
#include "ONNXEngine.h"
|
|
#include <vector>
|
|
#include <string>
|
|
#include <mutex>
|
|
|
|
namespace ANSCENTER {
|
|
namespace onnxocr {
|
|
|
|
class ONNXOCRRecognizer : public BasicOrtHandler {
|
|
public:
|
|
explicit ONNXOCRRecognizer(const std::string& onnx_path, unsigned int num_threads = 1);
|
|
~ONNXOCRRecognizer() override = default;
|
|
|
|
// Load character dictionary (must be called before Recognize)
|
|
bool LoadDictionary(const std::string& dictPath);
|
|
|
|
// Recognize text from a single cropped text image
|
|
TextLine Recognize(const cv::Mat& croppedImage);
|
|
|
|
// Batch recognition for multiple cropped images
|
|
std::vector<TextLine> RecognizeBatch(const std::vector<cv::Mat>& croppedImages);
|
|
|
|
private:
|
|
Ort::Value transform(const cv::Mat& mat) override;
|
|
Ort::Value transformBatch(const std::vector<cv::Mat>& images) override;
|
|
|
|
// CTC greedy decode
|
|
TextLine CTCDecode(const float* outputData, int seqLen, int numClasses);
|
|
|
|
std::vector<std::string> keys_;
|
|
int imgH_ = kRecImgH;
|
|
int imgMaxW_ = kRecImgMaxW;
|
|
std::mutex _mutex;
|
|
};
|
|
|
|
} // namespace onnxocr
|
|
} // namespace ANSCENTER
|