My-PaddleOCR
介紹
如何在 C++ 項目中,通過源碼使用 PaddlePaddle 實現 OCR 功能。
本項目的所有源碼:gitee: paddleocr
目前,官方提供使用 PaddleOcr 的方案有:
-
在 Python 項目中,調用
paddlepaddle
+paddleocr
包。 -
在 C++項目中,調用一個可執行文件。(由編譯 PaddleOCR 中的
deploy/cpp_infer
下的代碼形成)Paddle OCR 提供了一個通過編譯
deploy/cpp_infer
下的代碼爲ppocr.exe
,然後通過命令行調用獲取 OCR 的結果。
具體過程見: 服務器端 C++預測
其它方法:
- 使用 Python 寫一個 RESTful 服務,然後讓 C++項目調用這個服務功能。
這裏主要介紹一個更加直接的方法:
- 將
deploy/cpp_infer
的源碼引入到我們的 C++項目。 - 做適當修改後,我們就可以直接使用這些 API。
- 當然,如果你願意的話,也可以將這些源碼形成一個新的項目,編譯成一個 dll。(這裏並不介紹)
說明: - 下面的方法只在 release 版中有效。可能是因爲
paddle_inference.dll
只支持 release 版。
如果要支持 debug 版,可能需要重新編譯 PaddlePaddle。
Paddle OCR C++ 源碼
Paddle OCR 的倉庫,在github: PaddleOCR 或者 gitee: PaddleOCR
C++ 相關的代碼在目錄 deploy/cpp_infer
裏。
如何引入 Paddle OCR C++ 源碼
需要安裝的組件
- opencv
我在 opencv 4.6 版本上測試通過。
注意:opencv 4.5 版本存在一些問題,會導致功能異常。
設置環境變量:
@REM 設置opencv目錄環境變量
setx OPENCV_ROOT "C:\3rd\opencv4.6\build"
- paddle_inference
我用的是 2.6 版。
設置環境變量:
@REM 設置paddle_inference目錄環境變量
setx PADDLE_ROOT "C:\3rd\paddle_inference"
- cuda (optional)
你應該懂得。 - cudnn (optional)
你應該懂得。
創建 C++項目
-
創建一個空的 C++ 項目。
-
在項目中創建文件:
OpenCVDebug.props
<?xml version="1.0" encoding="utf-8"?>
<Project ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
<ImportGroup Label="PropertySheets" />
<PropertyGroup Label="UserMacros" />
<PropertyGroup />
<ItemDefinitionGroup>
<ClCompile>
<AdditionalIncludeDirectories>$(OPENCV_ROOT)\include;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
</ClCompile>
<Link>
<AdditionalLibraryDirectories>$(OPENCV_ROOT)\x64\vc15\lib;%(AdditionalLibraryDirectories)</AdditionalLibraryDirectories>
<AdditionalDependencies>opencv_world460d.lib;%(AdditionalDependencies)</AdditionalDependencies>
</Link>
</ItemDefinitionGroup>
<ItemGroup />
</Project>
- 在項目中創建文件:
OpenCVRelease.props
<?xml version="1.0" encoding="utf-8"?>
<Project ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
<ImportGroup Label="PropertySheets" />
<PropertyGroup Label="UserMacros" />
<PropertyGroup />
<ItemDefinitionGroup>
<ClCompile>
<AdditionalIncludeDirectories>$(OPENCV_ROOT)\include;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
</ClCompile>
<Link>
<AdditionalLibraryDirectories>$(OPENCV_ROOT)\x64\vc15\lib;%(AdditionalLibraryDirectories)</AdditionalLibraryDirectories>
<AdditionalDependencies>opencv_world460.lib;%(AdditionalDependencies)</AdditionalDependencies>
</Link>
</ItemDefinitionGroup>
<ItemGroup />
</Project>
- 在項目中創建文件:
PaddleDebug.props
<?xml version="1.0" encoding="utf-8"?>
<Project ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
<ImportGroup Label="PropertySheets" />
<PropertyGroup Label="UserMacros" />
<PropertyGroup />
<ItemDefinitionGroup>
<ClCompile>
<AdditionalIncludeDirectories>./;$(PADDLE_ROOT)/paddle/include;$(PADDLE_ROOT)/third_party/install/zlib/include;$(PADDLE_ROOT)/third_party/boost;$(PADDLE_ROOT)/third_party/eigen3;$(PADDLE_ROOT)/third_party/install/mklml/include;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
</ClCompile>
<Link>
<AdditionalLibraryDirectories>$(PADDLE_ROOT)/third_party/install/zlib/lib;$(PADDLE_ROOT)/third_party/install/mklml/lib;$(PADDLE_ROOT)/paddle/lib;%(AdditionalLibraryDirectories)</AdditionalLibraryDirectories>
<AdditionalDependencies>paddle_inference.lib;mklml.lib;libiomp5md.lib;%(AdditionalDependencies)</AdditionalDependencies>
</Link>
</ItemDefinitionGroup>
<ItemGroup />
</Project>
- 在項目中創建文件:
PaddleRelease.props
<?xml version="1.0" encoding="utf-8"?>
<Project ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
<ImportGroup Label="PropertySheets" />
<PropertyGroup Label="UserMacros" />
<PropertyGroup />
<ItemDefinitionGroup>
<ClCompile>
<AdditionalIncludeDirectories>./;$(PADDLE_ROOT)/paddle/include;$(PADDLE_ROOT)/third_party/install/zlib/include;$(PADDLE_ROOT)/third_party/boost;$(PADDLE_ROOT)/third_party/eigen3;$(PADDLE_ROOT)/third_party/install/mklml/include;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
</ClCompile>
<Link>
<AdditionalLibraryDirectories>$(PADDLE_ROOT)/third_party/install/zlib/lib;$(PADDLE_ROOT)/third_party/install/mklml/lib;$(PADDLE_ROOT)/paddle/lib;%(AdditionalLibraryDirectories)</AdditionalLibraryDirectories>
<AdditionalDependencies>paddle_inference.lib;mklml.lib;libiomp5md.lib;%(AdditionalDependencies)</AdditionalDependencies>
</Link>
</ItemDefinitionGroup>
<ItemGroup />
</Project>
- 修改 C++項目文件:
my-paddleocr.vcxproj
<ImportGroup Label="PropertySheets" Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">
<Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
<Import Project=".\OpenCVDebug.props" />
<Import Project=".\PaddleDebug.props" />
</ImportGroup>
<ImportGroup Label="PropertySheets" Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">
<Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
<Import Project=".\OpenCVRelease.props" />
<Import Project=".\PaddleRelease.props" />
</ImportGroup>
<ImportGroup Label="PropertySheets" Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">
<Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
<Import Project=".\OpenCVDebug.props" />
<Import Project=".\PaddleDebug.props" />
</ImportGroup>
<ImportGroup Label="PropertySheets" Condition="'$(Configuration)|$(Platform)'=='Release|x64'">
<Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
<Import Project=".\OpenCVRelease.props" />
<Import Project=".\PaddleRelease.props" />
</ImportGroup>
引入 Paddle OCR C++ 源碼
- 克隆 Paddle OCR 倉庫到本地。
- 切換到分支
release/2.6
。
git clone https://gitee.com/paddlepaddle/PaddleOCR.git
cd PaddleOCR
git checkout release/2.6
- 將
deploy/cpp_infer
目錄下的include
、src
兩個目錄的內容複製到我們的 C++ 項目中。 - 修改新的
src
目錄名稱爲ocr
。 - 刪除
ocr/main.cpp
- 在
include
目錄下創建ocr_flags.h
文件,內容如下:
這個文件是爲了替換 google 的 gflags 庫的使用。
#pragma once
// common args
#include <string>
using std::string;
#define DECLARE_bool(name) extern bool FLAGS_##name;
#define DECLARE_int32(name) extern int FLAGS_##name;
#define DECLARE_string(name) extern string FLAGS_##name;
#define DECLARE_double(name) extern double FLAGS_##name;
#define DEFINE_VARIABLE(type, name, value, help) \
static const type FLAGS_nono##name = value; \
type FLAGS_##name = FLAGS_nono##name; \
static type FLAGS_no##name = FLAGS_nono##name;
#define DEFINE_bool(name, val, txt) DEFINE_VARIABLE(bool, name, val, txt)
#define DEFINE_int32(name, val, txt) DEFINE_VARIABLE(int, name, val, txt)
#define DEFINE_string(name, val, txt) DEFINE_VARIABLE(string, name, val, txt)
#define DEFINE_double(name, val, txt) DEFINE_VARIABLE(double, name, val, txt)
- 修改
include/args.h
文件,內容如下:
不使用 gflags/gflags.h
註釋行 // #include <gflags/gflags.h>
增加行 #include "ocr_flags.h"
...
// #include <gflags/gflags.h>
#include "ocr_flags.h"
...
- 修改
ocr/args.cpp
文件,內容如下:
不使用 gflags/gflags.h
註釋行 // #include <gflags/gflags.h>
增加行 #include "include/ocr_flags.h"
...
// #include <gflags/gflags.h>
#include "include/ocr_flags.h"
...
- 修改
include/paddleocr.cpp
文件,內容如下:
不使用 auto_log/autolog.h
註釋行 // #include "auto_log/autolog.h"
註釋方法 PPOCR::benchmark_log
的內容。
// #include "auto_log/autolog.h"
...
void PPOCR::benchmark_log(int img_num) {
// if (this->time_info_det[0] + this->time_info_det[1] +
// this->time_info_det[2] >
// 0) {
// AutoLogger autolog_det("ocr_det", FLAGS_use_gpu, FLAGS_use_tensorrt,
// FLAGS_enable_mkldnn, FLAGS_cpu_threads, 1,
// "dynamic", FLAGS_precision, this->time_info_det,
// img_num);
// autolog_det.report();
// }
// if (this->time_info_rec[0] + this->time_info_rec[1] +
// this->time_info_rec[2] >
// 0) {
// AutoLogger autolog_rec("ocr_rec", FLAGS_use_gpu, FLAGS_use_tensorrt,
// FLAGS_enable_mkldnn, FLAGS_cpu_threads,
// FLAGS_rec_batch_num, "dynamic", FLAGS_precision,
// this->time_info_rec, img_num);
// autolog_rec.report();
// }
// if (this->time_info_cls[0] + this->time_info_cls[1] +
// this->time_info_cls[2] >
// 0) {
// AutoLogger autolog_cls("ocr_cls", FLAGS_use_gpu, FLAGS_use_tensorrt,
// FLAGS_enable_mkldnn, FLAGS_cpu_threads,
// FLAGS_cls_batch_num, "dynamic", FLAGS_precision,
// this->time_info_cls, img_num);
// autolog_cls.report();
// }
}
...
- 修改
include/paddlestructure.cpp
文件,內容如下:
不使用 auto_log/autolog.h
註釋行 // #include "auto_log/autolog.h"
註釋方法 PPOCR::benchmark_log
的內容。
// #include "auto_log/autolog.h"
...
void PaddleStructure::benchmark_log(int img_num) {
// if (this->time_info_det[0] + this->time_info_det[1] +
// this->time_info_det[2] >
// 0) {
// AutoLogger autolog_det("ocr_det", FLAGS_use_gpu, FLAGS_use_tensorrt,
// FLAGS_enable_mkldnn, FLAGS_cpu_threads, 1,
// "dynamic", FLAGS_precision, this->time_info_det,
// img_num);
// autolog_det.report();
// }
// if (this->time_info_rec[0] + this->time_info_rec[1] +
// this->time_info_rec[2] >
// 0) {
// AutoLogger autolog_rec("ocr_rec", FLAGS_use_gpu, FLAGS_use_tensorrt,
// FLAGS_enable_mkldnn, FLAGS_cpu_threads,
// FLAGS_rec_batch_num, "dynamic", FLAGS_precision,
// this->time_info_rec, img_num);
// autolog_rec.report();
// }
// if (this->time_info_cls[0] + this->time_info_cls[1] +
// this->time_info_cls[2] >
// 0) {
// AutoLogger autolog_cls("ocr_cls", FLAGS_use_gpu, FLAGS_use_tensorrt,
// FLAGS_enable_mkldnn, FLAGS_cpu_threads,
// FLAGS_cls_batch_num, "dynamic", FLAGS_precision,
// this->time_info_cls, img_num);
// autolog_cls.report();
// }
// if (this->time_info_table[0] + this->time_info_table[1] +
// this->time_info_table[2] >
// 0) {
// AutoLogger autolog_table("table", FLAGS_use_gpu, FLAGS_use_tensorrt,
// FLAGS_enable_mkldnn, FLAGS_cpu_threads,
// FLAGS_cls_batch_num, "dynamic", FLAGS_precision,
// this->time_info_table, img_num);
// autolog_table.report();
// }
// if (this->time_info_layout[0] + this->time_info_layout[1] +
// this->time_info_layout[2] >
// 0) {
// AutoLogger autolog_layout("layout", FLAGS_use_gpu, FLAGS_use_tensorrt,
// FLAGS_enable_mkldnn, FLAGS_cpu_threads,
// FLAGS_cls_batch_num, "dynamic",
// FLAGS_precision, this->time_info_layout,
// img_num);
// autolog_layout.report();
// }
}
...
- 修改
include/utility.cpp
文件,內容如下:
不使用 dirent.h
註釋行 // #include <dirent.h>
註釋方法 Utility::GetAllFiles
的內容。
// #include <dirent.h>
...
// list all files under a directory
void Utility::GetAllFiles(const char *dir_name,
std::vector<std::string> &all_inputs) {
// if (NULL == dir_name) {
// std::cout << " dir_name is null ! " << std::endl;
// return;
// }
// struct stat s;
// stat(dir_name, &s);
// if (!S_ISDIR(s.st_mode)) {
// std::cout << "dir_name is not a valid directory !" << std::endl;
// all_inputs.push_back(dir_name);
// return;
// } else {
// struct dirent *filename; // return value for readdir()
// DIR *dir; // return value for opendir()
// dir = opendir(dir_name);
// if (NULL == dir) {
// std::cout << "Can not open dir " << dir_name << std::endl;
// return;
// }
// std::cout << "Successfully opened the dir !" << std::endl;
// while ((filename = readdir(dir)) != NULL) {
// if (strcmp(filename->d_name, ".") == 0 ||
// strcmp(filename->d_name, "..") == 0)
// continue;
// // img_dir + std::string("/") + all_inputs[0];
// all_inputs.push_back(dir_name + std::string("/") +
// std::string(filename->d_name));
// }
// }
}
...
編寫調用程序
- 在項目中創建文件:
paddle_util.h
內容如下:
#pragma once
using namespace std;
#include <include/paddleocr.h>
class PaddleUtil {
public:
PaddleUtil();
public:
static PaddleUtil &get();
static void init();
void rec_image(const string &imageFile);
private:
PaddleOCR::PPOCR ocr;
};
- 在項目中創建文件:
paddle_util.cpp
內容如下:
#include "opencv2/core.hpp"
#include "opencv2/imgcodecs.hpp"
#include "opencv2/imgproc.hpp"
#include <iostream>
#include <vector>
#include "paddle_util.h"
#include <include/args.h>
using namespace PaddleOCR;
PaddleUtil::PaddleUtil() {}
PaddleUtil &PaddleUtil::get() {
static PaddleUtil self;
return self;
}
void PaddleUtil::init() {
FLAGS_det = true;
FLAGS_rec = true;
FLAGS_cls = false;
FLAGS_use_angle_cls = false;
FLAGS_det_model_dir = "model/whl/det/en/en_PP-OCRv3_det_infer";
FLAGS_rec_model_dir = "model/whl/rec/en/en_PP-OCRv4_rec_infer";
FLAGS_rec_char_dict_path = "model/en_dict.txt";
}
void PaddleUtil::rec_image(const string &imageFile) {
if (FLAGS_benchmark) {
ocr.reset_timer();
}
cv::Mat img = cv::imread(imageFile, cv::IMREAD_COLOR);
if (!img.data) {
std::cerr << "[ERROR] image read failed! image path: " << imageFile
<< std::endl;
return;
}
std::vector<OCRPredictResult> ocr_result =
ocr.ocr(img, FLAGS_det, FLAGS_rec, FLAGS_cls);
Utility::print_result(ocr_result);
if (FLAGS_visualize && FLAGS_det) {
std::string file_name = Utility::basename(imageFile);
Utility::VisualizeBboxes(img, ocr_result, FLAGS_output + "/" + file_name);
}
}
- 在項目中創建主程序文件:
my-paddleocr.cpp
內容如下:
#include <iostream>
#include "paddle_util.h"
int main() {
PaddleUtil::init();
PaddleUtil::get().rec_image("model/254.jpg");
std::cout << "Done!\n";
}
編譯運行
- 在 Visual Studio 中使用 Release 版本編譯。
- 運行
copy_paddle_dll.bat
,複製paddle_inference.dll
到輸出目錄。
@REM copy dll to Release
copy /Y %PADDLE_ROOT%\paddle\lib\paddle_inference.dll .\x64\Release\
copy /Y %PADDLE_ROOT%\third_party\install\mkldnn\lib\mkldnn.dll .\x64\Release\
copy /Y %PADDLE_ROOT%\third_party\install\mklml\lib\mklml.dll .\x64\Release\
copy /Y %PADDLE_ROOT%\third_party\install\mklml\lib\libiomp5md.dll .\x64\Release\
copy /Y %PADDLE_ROOT%\third_party\install\paddle2onnx\lib\paddle2onnx.dll .\x64\Release\
copy /Y %PADDLE_ROOT%\third_party\install\onnxruntime\lib\onnxruntime.dll .\x64\Release\