Getting Started

OCR SDK

OCR SDK provides OCR technology as an SDK that recognizes and detects text contained in image files and extracts it as data. Powered by proprietary deep learning technology, it accurately recognizes printed text, handwriting, and text within background images.

🔍 Precise recognition

🧠 Deep learning-based

Quick start guide

1. Check system requirements

Windows

Windows system requirements

• Operating system: Windows 10 or later, 64-bit required
• Compiler: A compiler that supports C++17 or later
• Processor: i7-8700K or higher recommended
• Memory: 8 GB RAM or more (16 GB recommended)
• Storage: 4 GB or more (SSD recommended)
• Other: Microsoft Visual C++ 2015–2022 Redistributable required

Detailed system requirements

See Supported operating systems for details.

2. License activation

License activation proceeds as follows.

Detailed license guide

See the License guide for license types and management.

3. Install prerequisites

Common prerequisites

• Libtorch library: Required to use the OCR SDK
• Encoding settings: Use an appropriate encoding such as UTF-8

Detailed installation guide

See Common prerequisites for installation instructions.

4. SDK initialization methods

IMR file

Initialize with an integrated resource file

// Initialize with an IMR file
void *obj = nullptr;
std::string runtimeLicensePath = "../bin64/Hancom.trial.license";
std::string engineLibPath = "../bin64/libs/CPU/modulelibs";
std::string imrResourcePath = "../assets/Hancom.25.12.01.cpu.imr";
std::string objName = "Sample";
bool isRecovery = true;
bool isSelfCheck = false;
int enginesCount = 1;

bool isSuccess = HOCR_SDK_CPU_CreateFromImrModelFile(
    &obj,
    runtimeLicensePath.data(),
    engineLibPath.data(),
    imrResourcePath.data(),
    objName.data(),
    isRecovery,
    isSelfCheck,
    enginesCount
);

Step-by-step initialization

Step-by-step initialization

// Create the SDK object and then initialize
std::string runtimeLicensePath = "../bin64/Hancom.trial.license";
std::string engineLibPath = "../bin64/libs/CPU/modulelibs";
std::string logFilePath = "./logs";

void* obj = nullptr;
obj = HOCR_SDK_CPU_CreateArgoSDK(
    runtimeLicensePath.data(),
    engineLibPath.data(),
    logFilePath.data()
);

std::string imrResourcePath = "../assets/Hancom.25.12.01.cpu.imr";
std::string objName = "Sample";
bool isRecovery = true;
bool isSelfCheck = false;
int enginesCount = 1;

bool isSuccess = HOCR_SDK_CPU_initFromImrModelFile(
    obj,
    imrResourcePath.data(),
    objName.data(),
    isRecovery,
    isSelfCheck,
    enginesCount
);

5. Basic OCR processing example

For image processing, OpenCV is used as an external library example.

/* ... */
#include "HOCR_SDK_COMMON.h"
#include "HOCR_SDK_CPU.h"

int main() {
    // Variables required for initialization
    void* obj = nullptr;
    std::string licensePath = ".\\Hancom.trial.license";
    std::string engineLibPath = ".\\libs\\CPU\\modulelibs";
    std::string imrModelDataPath = "..\\assets\\Hancom.25.11.18.cpu.imr";
    std::string objName = "Sample";
    bool isRecovery = true;
    bool isSelfCheck = false;
    int engineCount = 1;

    // Initialize SDK
    bool isSuccess = HOCR_SDK_CPU_CreateFromImrModelFile(
        &obj,
        licensePath.data(),
        engineLibPath.data(),
        imrModelDataPath.data(),
        objName.data(),
        isRecovery,
        isSelfCheck,
        engineCount
    );
    if (!isSuccess) return 1;

    // OCR test
    cv::Mat mat = cv::imread("ocr1.png", cv::IMREAD_COLOR);
    char *resultJsonStr = nullptr;
    std::string openApiName = "TEXT_RECOGNITION";
    std::string requestId = "request-id";
    char *dataMat = reinterpret_cast<char*>(temp.data);
    std::string requestOptionJsonStr = "{}";
    int resultSize = HOCR_SDK_run(
      obj,
      &resultJsonStr,
      openApiName.data(),
      requestId.data(),
      mat.cols,
      mat.rows,
      2, /* RGB 0, RGBA 1, BGR 2, BGRA 3, YUV 4, GRAY 5 */
      dataMat,
      requestOptionJsonStr.data()
    );
    std::string resultJson(resultJsonStr, resultSize);
    std::cout << resultJson << std::endl;

    // Release SDK
    HOCR_SDK_freePointer(&resultJsonStr);
    isSuccess = HOCR_SDK_releaseArgoSDK(obj);
    if (!isSuccess) return 1;

    return 0;
}

Sample program

You can test the SampleWindowsMain.exe sample program that uses the example above. See OCR SDK details.

SDK components

Distribution components

HANCOM OCRSDK
├─assets (OCRSDK resource directory)
│
├─bin64 (OCRSDK binary directory)
│  │
│  └─libs
│      └─CPU
│          └─modulelibs (OCRSDK module binary directory)
│
├─include (OCRSDK header file directory)
│
└─notice (OCRSDK open-source license notice directory)

Detailed components

See all components in OCR SDK details.