Win10 环境下 VS2022 暴力编译PP-OCRv4

1 环境准备

1. 下载PaddleOCR
   PaddleOCR
   C++ 部署代码位于 PaddleOCR\deploy\cpp_infer目录下
   
   复制cpp_infer目录下include和src到项目目录下
2. paddle_inference
   paddle_inference
   
3. opencv
   这里使用已经安装好的opencv4.5.5
4. 下载dirent-master.zip
   下载dirent-master.zip, 解压并复制dirent.h文件到项目目录下
5. 下载权重文件
   检测模型 ch_PP-OCRv4_det_infer
   方向分类器 ch_ppocr_mobile_v2.0_cls_infer
   识别模型 ch_PP-OCRv4_rec_infer

2 Visual Studio 2022 环境配置

属性 --> VC++ 目录 --> 包含目录

E:\opencv455\build\include
E:\cpp_code\PaddleOCR\paddle_inference\paddle\include
E:\cpp_code\PaddleOCR\paddle_inference\third_party\install\cryptopp\include
E:\cpp_code\PaddleOCR\paddle_inference\third_party\install\protobuf\include
E:\cpp_code\PaddleOCR\paddle_inference\third_party\install\glog\include
E:\cpp_code\PaddleOCR\paddle_inference\third_party\install\gflags\include
E:\cpp_code\PaddleOCR\paddle_inference\third_party\install\xxhash\include
E:\cpp_code\PaddleOCR\paddle_inference\third_party\install\onnxruntime\include
E:\cpp_code\PaddleOCR\paddle_inference\third_party\install\paddle2onnx\include
E:\cpp_code\PaddleOCR\paddle_inference\third_party\install\mklml\include
E:\cpp_code\PaddleOCR\paddle_inference\third_party\install\mkldnn\include
E:\cpp_code\PaddleOCR\paddle_inference\third_party\install\utf8proc\include
E:\cpp_code\PaddleOCR\ocr_cpu
1
2
3
4
5
6
7
8
9
10
11
12
13

属性 --> VC++ 目录 --> 库目录

E:\opencv455\build\x64\vc15\lib
E:\cpp_code\PaddleOCR\paddle_inference\paddle\lib
E:\cpp_code\PaddleOCR\paddle_inference\third_party\install\cryptopp\lib
E:\cpp_code\PaddleOCR\paddle_inference\third_party\install\protobuf\lib
E:\cpp_code\PaddleOCR\paddle_inference\third_party\install\glog\lib
E:\cpp_code\PaddleOCR\paddle_inference\third_party\install\gflags\lib
E:\cpp_code\PaddleOCR\paddle_inference\third_party\install\xxhash\lib
E:\cpp_code\PaddleOCR\paddle_inference\third_party\install\onnxruntime\lib
E:\cpp_code\PaddleOCR\paddle_inference\third_party\install\paddle2onnx\lib
E:\cpp_code\PaddleOCR\paddle_inference\third_party\install\mklml\lib
E:\cpp_code\PaddleOCR\paddle_inference\third_party\install\mkldnn\lib
E:\cpp_code\PaddleOCR\paddle_inference\third_party\install\utf8proc\lib
1
2
3
4
5
6
7
8
9
10
11
12

属性 --> 链接器 --> 输入 --> 附加依赖项

opencv_world455.lib
paddle_inference.lib
libpaddle_inference.lib
cryptopp-static.lib
gflags_static.lib
shlwapi.lib
glog.lib
mkldnn.lib
mklml.lib
libiomp5md.lib
onnxruntime.lib
paddle2onnx.lib
libprotobuf.lib
utf8proc_static.lib
xxhash.lib
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

将 opencv_world455.dll，paddle_inference.dll，paddle2onnx.dll 和 onnxruntime.dll复制到项目的x64\Release目录下

3 测试

修改 main.cpp文件

int main() {
    SetConsoleOutputCP(CP_UTF8);

    FLAGS_det_model_dir = "E:\\cpp_code\\weights\\PP-OCRv4\\ch_PP-OCRv4_det_infer";
    FLAGS_rec_model_dir = "E:\\cpp_code\\weights\\PP-OCRv4\\ch_PP-OCRv4_rec_infer";
    FLAGS_cls_model_dir = "E:\\cpp_code\\weights\\PP-OCRv4\\ch_ppocr_mobile_v2.0_cls_infer";
    FLAGS_rec_char_dict_path = "E:\\cpp_code\\paddle_compile\\PaddleOCR\\ppocr\\utils\\ppocr_keys_v1.txt";
    FLAGS_image_dir = "E:\\cpp_code\\images\\img_20230926_105631025.jpg";

    FLAGS_cpu_threads = 64;
    FLAGS_type = "ocr"; // "structure"
    FLAGS_output = ".\\output";


    std::vector<cv::String> cv_all_img_names;
    cv::glob(FLAGS_image_dir, cv_all_img_names);
    std::cout << "total images num: " << cv_all_img_names.size() << std::endl;

    if (FLAGS_type == "ocr") {
        ocr(cv_all_img_names);
    }
    else if (FLAGS_type == "structure") {
        structure(cv_all_img_names);
    }
    else {
        std::cout << "only value in ['ocr','structure'] is supported" << std::endl;
    }
}
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28

遇到的问题和解决方法：

1. error LNK2038: 检测到“RuntimeLibrary”的不匹配项: 值“MT_StaticRelease”不匹配值“MD_DynamicRelease”

1>gflags_static.lib(gflags.cc.obj) : error LNK2038: 检测到“RuntimeLibrary”的不匹配项: 值“MT_StaticRelease”不匹配值“MD_DynamicRelease”(ocr_cpu.obj 中)
1>gflags_static.lib(gflags_reporting.cc.obj) : error LNK2038: 检测到“RuntimeLibrary”的不匹配项: 值“MT_StaticRelease”不匹配值“MD_DynamicRelease”(ocr_cpu.obj 中)
1>gflags_static.lib(gflags_completions.cc.obj) : error LNK2038: 检测到“RuntimeLibrary”的不匹配项: 值“MT_StaticRelease”不匹配值“MD_DynamicRelease”(ocr_cpu.obj 中)
1>libcpmt.lib(locale0.obj) : error LNK2038: 检测到“RuntimeLibrary”的不匹配项: 值“MT_StaticRelease”不匹配值“MD_DynamicRelease”(ocr_cpu.obj 中)
1>libcpmt.lib(cerr.obj) : error LNK2038: 检测到“RuntimeLibrary”的不匹配项: 值“MT_StaticRelease”不匹配值“MD_DynamicRelease”(ocr_cpu.obj 中)
1>libcpmt.lib(locale.obj) : error LNK2038: 检测到“RuntimeLibrary”的不匹配项: 值“MT_StaticRelease”不匹配值“MD_DynamicRelease”(ocr_cpu.obj 中)
1>libcpmt.lib(iosptrs.obj) : error LNK2038: 检测到“RuntimeLibrary”的不匹配项: 值“MT_StaticRelease”不匹配值“MD_DynamicRelease”(ocr_cpu.obj 中)
1
2
3
4
5
6
7

解决方法： 属性 -> C/C++ -> 代码生成 -> 运行库

2. error LNK2001: 无法解析的外部符号 “__declspec(dllimport) public: class std::basic_ostream<char,struct std::char_traits > & __cdecl google::LogMessage::stream(void)” (_imp?stream@LogMessage@google@@QEAAAEAV?$basi{c}_{o}stream@DU?$char_traits@D@std@@@std@@XZ)

1>paddleocr.obj : error LNK2001: 无法解析的外部符号 "__declspec(dllimport) public: class std::basic_ostream<char,struct std::char_traits<char> > & __cdecl google::LogMessage::stream(void)" (__imp_?stream@LogMessage@google@@QEAAAEAV?$basic_ostream@DU?$char_traits@D@std@@@std@@XZ)
1>paddleocr.obj : error LNK2001: 无法解析的外部符号 "__declspec(dllimport) public: __cdecl google::LogMessage::~LogMessage(void)" (__imp_??1LogMessage@google@@QEAA@XZ)
1>paddleocr.obj : error LNK2001: 无法解析的外部符号 "__declspec(dllimport) public: __cdecl google::LogMessage::LogMessage(char const *,int)" (__imp_??0LogMessage@google@@QEAA@PEBDH@Z)
1>paddleocr.obj : error LNK2001: 无法解析的外部符号 "__declspec(dllimport) public: class std::basic_ostream<char,struct std::char_traits<char> > & __cdecl std::basic_ostream<char,struct std::char_traits<char> >::operator<<(bool)" (__imp_??6?$basic_ostream@DU?$char_traits@D@std@@@std@@QEAAAEAV01@_N@Z)
1
2
3
4

解决方法： 属性 ->C/C++ ->预处理器 ->预处理器定义：GOOGLE_GLOG_DLL_DECL=

3. gflags_static.lib(gflags.cc.obj) : error LNK2001: 无法解析的外部符号 __imp_PathMatchSpecA

解决方法：项目属性 -》链接器-》输入-》附加依赖项-》加上shlwapi.lib