Coding is the most interesting phase of the Software Development. By the time we get here we must already know the below items to proceed:

1. The problem definition, we must know exactly what the problem is. In our hobby project we are exploring i.e. Facial Features Recognizer we know we want to recognize the facial features from people’s faces in the video and print it on the same video frames.
2. The initial solution to the problem, i.e. the result from the design phase. I get the visuals of imaginary code in my mind when scribbling down my design in paper.

At this point the rush to write code is at its peak because I might forget my solution. And it will be very difficult to get it back. So let’s start coding.

Project Structure

First we create the project folder structure or simply project structure. This is necessary because we will be writing a minimum of four classes and a main function. And also the code for unit testing. So totally we will be writing one kilo lines of code by blind guess backed by my work experience. Ya, its called one kilo lines instead of one thousand lines. In case you dont understand any part of my tutorial please follow my instructions blindly and do your experiments and try to learn on your own. As coding cannot be taught step-by-step and cannot be explained line-by-line.

We create the project structure from the command line using the terminal. Because we are hardcore programmers and terminal is our everything. So we start from here, we create the project structure as explained in my first blog Hello World C++ CMake project for Eclipse IDE. We make use of the tools I have written about in my second blog About C++ Tools. We arrive at the below project structure.

Project Structure

Note: This project structure is that of the completed one. In coding stage we just have to create inc, src, and CMakeLists.txt. We create the rest later and as needed.

Initial code

Tools	Version
Ubuntu	18.04 (LTS)
CPU	x86_64
CMake	3.10.2
G++	7.3.0
Eclipse	2018-09
Languages	C++, CMake

I always begin my project by writing the minimum code required to print a simple statement. As we are going to use many components here such as cmake, opencv and google tests library, writing the minimal code and testing its execution ensures that the required component works as expected.

First we write the cmake script to set the project name and add a target to build the project executable. The cmake script should be placed at the root of the project structure i.e. in FacialFeaturesRecognizer folder. The cmake code for the same is written in CMakeLists.txt as shown below.

cmake_minimum_required(VERSION 2.8)
set (CMAKE_CXX_STANDARD 11)

project(FacialFeaturesRecognizer)
get_filename_component(MODULE_NAME ${CMAKE_CURRENT_SOURCE_DIR} NAME)

#-- target for exe
include_directories(${CMAKE_CURRENT_SOURCE_DIR}/inc/)
file(GLOB module_srcs ${CMAKE_CURRENT_SOURCE_DIR}/src/*.C++)
set(SRCS main.C++ ${module_srcs})

string(TOLOWER ${MODULE_NAME} MODULE_NAME)
add_executable(${MODULE_NAME} ${SRCS})

Initial cmake code for the project

Write the main function (the entry point to the executable) in a separate file such as main.C++. Let’s place the main.C++ file in the project root folder i.e. in FacialFeaturesRecognizer folder. By separating the main function from the application code we can add another target to build the same code as a library to be used in some other scenario.

#include <BaseRecognizer.h>

using namespace FFR;

int main(int argc, char **argv) {
  return execute(argc, argv);
}  // end

Entry point to the executable

Create inc and src the folders to save the project header files and source files in respectively.

Generate Eclipse IDE Project

Now we use the cmake script to generate the eclipse project to start coding in Eclipse IDE. Go to build folder and enter the below command in terminal to generate the eclipse project files.

cmake ../FacialFeaturesRecognizer \
       -G"Eclipse CDT4 - Unix Makefiles" \
       -DCMAKE_ECLIPSE_GENERATE_SOURCE_PROJECT=TRUE \
       -DCMAKE_BUILD_TYPE=Debug \
       -DCMAKE_CXX_FLAGS="-std=c++11" \
       -DCMAKE_CXX_COMPILER_ARG1=-std=c++11

Commands to generate the Eclipse IDE project

The eclipse project files .project and .cproject will be generated along with the cmake build files and stored in the build folder. When we import this folder into eclipse workspace as an existing eclipse project we can see the project details such as the build targets, binaries, source files and cmake scripts.

I m creating the eclipse project from terminal and importing it manually instead of asking eclipse to parse the cmake file and automatically generate the project files which gives me the advantage of being able to control the project from my terminal as the IDE hides many processes in working with the project. For example it hides compilation flags passed to the compiler.

There are few other ways to generate the eclipse project from the cmake, they are explained in the link to the official docs.

Coding in Eclipse IDE

Start the Eclipse IDE. Click File->Import->General->Existing Projects into Workspace. Browse to the folder containing the cmake generated eclipse project files for the our project i.e. build folder and import it. If the project is successfully imported then we can click Finish to continue editing the code in the Eclipse IDE.

The C++ project in Eclipse IDE looks like as shown below: Eclipse C++ Project

Legends:

1. Title of the project as we set in the CMakeLists.txt file. The postfix “-Debug” is appended to the project name to indicate that it is a debug build type.
2. The build target we added in the CMakeLists.txt file. Since we added only one target hence I have highlighted it. The other build targets are just utilities. We will use this target, all, and clean more commonly than the other targets. More targets will be added later for unit tests, documentations etc and they show up when re-import the project.
3. It is the link to the source directory i.e. FacialFeaturesRecognizer folder. Since the eclipse project description .project is stored in the build folder the cmake creates a link to the source folder. The changes we make in this linked folder also reflects in the actual source folder.

Let’s create the BaseRecognizer class with the a simple printf statement to confirm the working of the project setup. Right click [Source directory] and go to New->Class enter the name of the class i.e. BaseRecognizer and the namespace FFR (can be any unique identifer for the project). Prepend the header file name with inc/ and src/ this will create the header file for BaseRecognizer class in inc and src folder respectively. At the bottom there is also an option to add a unit test code for the class. We ignore the unit test code for now as we do not know what unit tests to write.

After creating the class, write the declarations in the header file and definitions in the source file as:

#ifndef INC_BASERECOGNIZER_H_
#define INC_BASERECOGNIZER_H_

#include <cstdio>

namespace FFR {

class BaseRecognizer {
public:
	BaseRecognizer();
	virtual ~BaseRecognizer();

	void printLog(void);
};

extern int execute(int argc, char **argv);

} /* namespace FFR */

#endif /* INC_BASERECOGNIZER_H_ */

BaseRecognizer.h

#include <BaseRecognizer.h>

namespace FFR {

BaseRecognizer::BaseRecognizer() {
	// Auto-generated constructor stub

}

BaseRecognizer::~BaseRecognizer() {
	// Auto-generated destructor stub
}

void BaseRecognizer::printLog() {
	printf("HelloWorld\n");
}

/**
 * Entry point into the application
 */
int execute(int argc, char **argv) {
	int err = 0;

	// Detect facial features in the default video stream
	BaseRecognizer br;
	br.printLog();

	return static_cast<int>(err);
}

} /* namespace FFR */

BaseRecognizer.cpp

Build the project from the eclipse by double clicking the target for the project. If it shows undefined reference error then delete the project from the Eclipse Workspace without deleting the contents from the disk. Run the same cmake command once again and re-import the project in eclipse. This is done because the cmake build files were not aware of the newly created source files.

Run the project by selecting Run As->Local C/C++ Application. The HelloWorld is printed in the console view of the Eclipse IDE.

Format code

We are ready to format the code to make it look professional.

1. Add the copyright header comment to indicate its creator, creation date, author and brief description about the contents in the file.
2. Click Ctrl+Shift+F to automatically format the source code in eclipse. I use the Google Code Style to format the code as it is very good and easy to read format. It does not come bundled with the Eclipse IDE. So download it from the link and import it in the Eclipse.

The code is made professional and it is ready for writing further coding.

Note: It will be a while before I write next blog post which explains full code because I have to uprgade the OpenCV library to 4.0 as I dont feel like writing a blog about working with a ten years old library.

The code hosted on my github account for this project is written using the OpenCV 2.4 with the outdated APIs for SVM and ML modules.