Version: 3.16.1 (latest)

Face Recognition in a Video Stream

In this tutorial you'll learn how to recognize faces in a video stream. For recognition, use a ready-made database of faces from Face SDK distribution package. The database includes the images of several famous people. Recognized faces are highlighted with a green rectangle. The name and image of a recognized person are displayed next to the person's face in a video stream. This tutorial is based on Face Detection and Tracking in a Video Stream and the corresponding project.

You can find the tutorial project in Face SDK: examples/tutorials/face_recognition_with_video_worker

Set Up the Project#

In Face Detection and Tracking in a Video Stream, we set only two parameters of Face SDK (a path to Face SDK and a configuration file name for the VideoWorker object). However, in this tutorial we need to set several more parameters: we'll add a path to the database, a configuration file name with a recognition method, and FAR. For convenience, we'll modify several files. Specify all the parameters in the FaceSdkParameters structure. In facesdkparameters.hspecify the path to the video_worker_lbf.xml configuration file.

facesdkparameters.h

struct FaceSdkParameters
{
    ...
    std::string videoworker_config = "video_worker_lbf.xml";
};

Pass the face_sdk_parameters structure to the constructor of the Worker object.

viewwindow.h

class ViewWindow : public QWidget
{
    Q_OBJECT

    public:
        explicit ViewWindow(
            QWidget *parent,
            pbio::FacerecService::Ptr service,
            FaceSdkParameters face_sdk_parameters);
    ...
    private:
    ...
        std::shared_ptr<Worker> _worker;

        pbio::FacerecService::Ptr _service;
};

viewwindow.cpp

ViewWindow::ViewWindow(
    QWidget *parent,
    pbio::FacerecService::Ptr service,
    FaceSdkParameters face_sdk_parameters) :
QWidget(parent),
ui(new Ui::ViewWindow),
_service(service)
{
    ui->setupUi(this);
    ...
    _worker = std::make_shared<Worker>(
        _service,
        face_sdk_parameters);
    ...
};

worker.h

#include "qcameracapture.h"
#include "facesdkparameters.h"
...
class Worker
{
    ...
    Worker(
        const pbio::FacerecService::Ptr service,
        const FaceSdkParameters face_sdk_parameters);
    ...
};

worker.cpp

Worker::Worker(
    const pbio::FacerecService::Ptr service,
    const FaceSdkParameters face_sdk_parameters)
{
    pbio::FacerecService::Config vwconfig(face_sdk_parameters.videoworker_config);
    ...
}

In this project we focus only on face detection in a video stream (creating a bounding rectangle) and face recognition. Note that in the first project (detection_and_tracking_with_video_worker), which you can use as a reference for this project, we also displayed anthropometric points and angles. If you don't want to display this information, you can just remove unnecessary visualization from the first project.

Create the Database of Faces#

First of all, create a database of faces. To check face recognition, you can use the ready-made database from Face SDK. This database includes images of three famous people (Elon Musk, Emilia Clarke, Lionel Messi). To check recognition, copy the database to the project root folder (next to a .pro file), run the project, open an image from the database, and point a camera at the screen. You can also add your picture to the database. To do this, create a new folder in the database, specify your name in a folder name, and copy your picture to the folder (in the same way as other folders in the database).
Create a new Database class to work with the database: Add New > C++ > C++ Class > Choose... > Class name – Database > Next > Project Management (default settings) > Finish. In database.h include the QImage and QString headers to work with images and strings and libfacerec.h to integrate Face SDK.

database.h

#include <QImage>
#include <QString>

#include <facerec/libfacerec.h>

class Database
{
    public:

    Database();
}

In database.cpp include the database.h and videoframe.h headers (implementation of the IRawImage interface, used by VideoWorker to receive the frames). Also, include necessary headers for working with the file system, debugging, exception handling, and working with files.

database.cpp

#include "database.h"
#include "videoframe.h"

#include <QDir>
#include <QDebug>

#include <stdexcept>
#include <fstream>

In database.h add a constructor and set the path to the database. Specify the Recognizer object to create templates, and specify the Capturer object to detect faces and far. FAR is frequency that the system makes false accepts. False accept means that a system claims a pair of pictures is a match, when these are actually the pictures of different individuals. The vw_elements vector contains the elements of the VideoWorker database. The thumbnails and names vectors contain the previews of images and names of people from the database.

database.h

class Database
{
    public:
    ...
    // Create a database
    Database(
        const std::string database_dir_path,
        pbio::Recognizer::Ptr recognizer,
        pbio::Capturer::Ptr capturer,
        const float fa_r);

    std::vector<pbio::VideoWorker::DatabaseElement> vw_elements;
    std::vector<QImage> thumbnails;
    std::vector<QString> names;
}

In database.cpp implement the Database constructor, declared in the previous subsection. The distance_threshold value means the recognition distance. Since this distance is different for different recognition methods, we get it based on the FAR value using the getROCCurvePointByFAR method.

database.cpp

Database::Database(
    const std::string database_dir_path,
    pbio::Recognizer::Ptr recognizer,
    pbio::Capturer::Ptr capturer,
    const float fa_r)
{
    const float distance_threshold = recognizer->getROCCurvePointByFAR(fa_r).distance;
}

Specify the path to the database with faces in the database_dir variable. If this path doesn't exist, you'll see the "database directory doesn't exist" exception. Create a new person_id variable to store the id of a person from the database (folder name in the database) and the element_id variable to store the id of an element in the database (a person's image from the database). In the dirs list, create a list of all subdirectories of the specified directory with the database.

database.cpp

Database::Database(
    const std::string database_dir_path,
    pbio::Recognizer::Ptr recognizer,
    pbio::Capturer::Ptr capturer,
    const float fa_r)
{
    ...

    QDir database_dir(QString::fromStdString(database_dir_path));

    if (!database_dir.exists())
    {
        throw std::runtime_error(database_dir_path + ": database directory doesn't exist");
    }

    int person_id = 0;
    int element_id_counter = 0;

    QFileInfoList dirs = database_dir.entryInfoList(
        QDir::AllDirs | QDir::NoDotAndDotDot,
        QDir::DirsFirst);
}

Note: See more information about FAR and TAR values for different recognition methods in Identification Performance.

In the for(const auto &dir: dirs) loop, process each subdirectory (data about each person). A folder name corresponds to a person's name. Create a list of images in person_files.

database.cpp

Database::Database(...)
{
    ...
    for(const auto &dir: dirs)
    {
        QDir person_dir(dir.filePath());

        QString name = dir.baseName();

        QFileInfoList person_files = person_dir.entryInfoList(QDir::Files | QDir::NoDotAndDotDot);
    }
}

In the for(const auto &person_file: person_files) nested loop, process each image. If an image doesn't exist, the "Can't read image" warning is displayed.

database.cpp

Database::Database(...)
{
    ...
    for(const auto &dir: dirs)
    {
        ...        
        QFileInfoList person_files = person_dir.entryInfoList(QDir::Files | QDir::NoDotAndDotDot);
        
        for(const auto &person_file: person_files)
        {
            QString path = person_file.filePath();

            qDebug() << "processing" << path << "name:" << name;

            QImage image(path);
            if(image.isNull())
            {
                qDebug() << "\n\nWARNING: cant read image" << path << "\n\n";
                continue;
            }

            if (image.format() != QImage::Format_RGB888)
            {
                image = image.convertToFormat(QImage::Format_RGB888);
            }

            VideoFrame frame;
            frame.frame() = QCameraCapture::FramePtr(new QImage(image));
        }
    }
}

Detect a face in an image using the Capturer object. If an image cannot be read, a face cannot be found in an image or more than one face is detected, the warning is displayed and this image is ignored.

database.cpp

Database::Database(...)
{
    ...
    for(const auto &dir: dirs)
    {
        ...        
        QFileInfoList person_files = person_dir.entryInfoList(QDir::Files | QDir::NoDotAndDotDot);
        
        for(const auto &person_file: person_files)
        {
            ...
            // Detect faces
            const std::vector<pbio::RawSample::Ptr> captured_samples = capturer->capture(frame);

            if(captured_samples.size() != 1)
            {
                qDebug() << "\n\nWARNING: detected" << captured_samples.size() <<
                    "faces on" << path << "image instead of one, image ignored\n\n";
                continue;
            }
            const pbio::RawSample::Ptr sample = captured_samples[0];
        }
    }
}

Using the recognizer->processing method, create a face template used for recognition.

database.cpp

Database::Database(...)
{
    ...
    for(const auto &dir: dirs)
    {
        ...        
        QFileInfoList person_files = person_dir.entryInfoList(QDir::Files | QDir::NoDotAndDotDot);
        
        for(const auto &person_file: person_files)
        {
            ...
            // Create a template
            const pbio::Template::Ptr templ = recognizer->processing(*sample);
        }
    }
}

In the pbio::VideoWorker::DatabaseElement vw_element structure, specify all the information about the database element that will be passed for processing to the VideoWorker object (element id, person id, face template, recognition threshold). Using the push_back method, add an element to the end of the list.

database.cpp

Database::Database(...)
{
    ...
    for(const auto &dir: dirs)
    {
        ...        
        QFileInfoList person_files = person_dir.entryInfoList(QDir::Files | QDir::NoDotAndDotDot);
        
        for(const auto &person_file: person_files)
        {
            ...
            // Prepare data for VideoWorker
            pbio::VideoWorker::DatabaseElement vw_element;
            vw_element.element_id = element_id_counter++;
            vw_element.person_id = person_id;
            vw_element.face_template = templ;
            vw_element.distance_threshold = distance_threshold;

            vw_elements.push_back(vw_element);
            thumbnails.push_back(makeThumbnail(image));
            names.push_back(name);
        }

        ++person_id;
    }
}

In database.h, add the makeThumbnail method to create a preview of a picture from the database.

database.cpp

class Database
{
    public:
        // Create a preview from a sample
        static
        QImage makeThumbnail(const QImage& image);
        ...
};

In database.cpp, implement the method using makeThumbnail to create a preview of a picture from the database, which will be displayed next to the face of a recognized person. Set the preview size (120 pixels) and scale it (keep the ratio if the image is resized).

database.cpp

#include <fstream>
...
QImage Database::makeThumbnail(const QImage& image)
{
    const float thumbnail_max_side_size = 120.f;

    const float scale = thumbnail_max_side_size / std::max<int>(image.width(), image.height());

    QImage result = image.scaled(
        image.width() * scale,
        image.height() * scale,
        Qt::KeepAspectRatio,
        Qt::SmoothTransformation);

    return result;
}

In the .pro file, set the path to the database.

face_recognition_with_video_worker.pro

...
DEFINES += FACE_SDK_PATH=\\\"$$FACE_SDK_PATH\\\"

DEFINES += DATABASE_PATH=\\\"$${_PRO_FILE_PWD_}/base\\\"

INCLUDEPATH += $${FACE_SDK_PATH}/include
...

In facesdkparameters.h, set the path to the database and the FAR value.

facesdkparameters.h

struct FaceSdkParameters
{
    ...
    std::string videoworker_config = "video_worker_lbf.xml";

    std::string database_dir = DATABASE_PATH;
    const float fa_r = 1e-5;
};

Search a Face in the Database and Display the Result#

In facesdkparameters.h set the path to the configuration file with the recognition method. In this project we use the method 6.7 because it is suitable for video stream processing and provides optimal recognition speed and good quality. You can learn more about recommended recognition methods in Face Identification.

facesdkparameters.h

struct FaceSdkParameters
{
    ...
    std::string videoworker_config = "video_worker_lbf.xml";

    std::string database_dir = DATABASE_PATH;
    const float fa_r = 1e-5;
    std::string method_config = "method6v7_recognizer.xml";
};

Note: If you want to recognize faces in a video stream using low-performance devices, you can use the method 9.30. In this case recognition speed is higher but recognition quality is lower compared to the method 6.7.

In worker.h add the match_database_index variable to the FaceData structure. This variable will store the database element, if a person is recognized, or "-1" if a person isn't recognized. Add Database and a callback indicating that a person is recognized (MatchFoundCallback).

worker.h

#include "qcameracapture.h"
#include "facesdkparameters.h"
#include "database.h"
...
class Worker
{
public:

    struct FaceData
    {
        pbio::RawSample::Ptr sample;
            bool lost;
            int frame_id;
            int match_database_index;

            FaceData() :
                lost(true),
                match_database_index(-1)
            {
            }
    };
    ...
    pbio::VideoWorker::Ptr _video_worker;

    Database _database;
    ...
    static void TrackingLostCallback(
        const pbio::VideoWorker::TrackingLostCallbackData &data,
        void* const userdata);

    static void MatchFoundCallback(
        const pbio::VideoWorker::MatchFoundCallbackData &data,
        void* const userdata);

    int _tracking_callback_id;
    int _tracking_lost_callback_id;
    int _match_found_callback_id;
};

In worker.cpp, override the parameter value in the configuration file so that MatchFoundCallback is received for non-recognized faces too. Set the parameters of the VideoWorker object: in the first tutorial we didn't recognize faces, that's why we set only the value of streams_count. Since in this project we'll recognize faces in a video stream, we need to specify in the constructor the path to the configuration file with the recognition method, and also the values of processing_threads_count (number of threads to create templates) and matching_threads_count (number of threads to compare the templates). In this project we use only one stream (a webcam connected to our PC). Connect the database: pass the path to the database, create Capturer to detect faces and Recognizer to create templates, and also specify the FAR coefficient. Using the setDatabase method, set the database for VideoWorker. Using the addMatchFoundCallback method, add the MatchFoundrecognition event handler.

worker.cpp

Worker::Worker(
    const pbio::FacerecService::Ptr service,
    const FaceSdkParameters face_sdk_parameters)
{
    pbio::FacerecService::Config vwconfig(face_sdk_parameters.videoworker_config);
   
    vwconfig.overrideParameter("not_found_match_found_callback", 1);

    _video_worker = service->createVideoWorker(
        vwconfig,
        face_sdk_parameters.method_config,
        1,   // streams_count
        1,   // processing_threads_count
        1);  // matching_threads_count

    _database = Database(
        face_sdk_parameters.database_dir,
        service->createRecognizer(face_sdk_parameters.method_config, true, false),
        service->createCapturer("common_capturer4_lbf_singleface.xml"),
        face_sdk_parameters.fa_r);

    _video_worker->setDatabase(_database.vw_elements);
    ...

    _match_found_callback_id =
        _video_worker->addMatchFoundCallbackU(
            MatchFoundCallback,
            this);
}

In the Worker::~Worker() destructor, remove MatchFoundCallback.

worker.cpp

Worker::~Worker()
{
    _video_worker->removeTrackingCallback(_tracking_callback_id);
    _video_worker->removeTrackingLostCallback(_tracking_lost_callback_id);
    _video_worker->removeMatchFoundCallback(_match_found_callback_id);
}
...

In MatchFoundCallback, the result is received in the form of the MatchFoundCallbackData structure that stores the information about recognized and unrecognized faces.

worker.cpp

// static
void Worker::TrackingLostCallback(
    const pbio::VideoWorker::TrackingLostCallbackData &data,
    void* const userdata)
{
    ...
}

// static
void Worker::MatchFoundCallback(
    const pbio::VideoWorker::MatchFoundCallbackData &data,
    void* const userdata)
{
    assert(userdata);

    const pbio::RawSample::Ptr &sample = data.sample;
    const pbio::Template::Ptr &templ = data.templ;
    const std::vector<pbio::VideoWorker::SearchResult> &search_results = data.search_results;

    // Information about a user - a pointer to Worker
    // Pass the pointer
    Worker &worker = *reinterpret_cast<Worker*>(userdata);

    assert(sample);
    assert(templ);
    assert(!search_results.empty());
}

When a template for a tracked person is created, this template is compared with each template from the database. If the distance to the closest element is less than distance_threshold specified in this element, then it is a match. If a face in a video stream is not recognized, then you'll see the message "Match not found". If a face is recognized, you'll see the message "Match found with..." and the name of the matched person.

worker.cpp

// static
void Worker::MatchFoundCallback(
    const pbio::VideoWorker::MatchFoundCallbackData &data,
    void* const userdata)
{
    ...    
    for(const auto &search_result: search_results)
    {
        const uint64_t element_id = search_result.element_id;

        if(element_id == pbio::VideoWorker::MATCH_NOT_FOUND_ID)
        {
            std::cout << "Match not found" << std::endl;
        }
        else
        {
            assert(element_id < worker._database.names.size());

            std::cout << "Match found with '"
                << worker._database.names[element_id].toStdString() << "'";
        }
    }
    std::cout << std::endl;
}

Save the data about the recognized face to display a preview.

worker.cpp

// static
void Worker::MatchFoundCallback(
    const pbio::VideoWorker::MatchFoundCallbackData &data,
    void* const userdata)
{
    ...
    const uint64_t element_id = search_results[0].element_id;

    if(element_id != pbio::VideoWorker::MATCH_NOT_FOUND_ID)
    {
        assert(element_id < worker._database.thumbnails.size());

        // Save the best matching result for rendering
        const std::lock_guard<std::mutex> guard(worker._drawing_data_mutex);

        FaceData &face = worker._drawing_data.faces[sample->getID()];

        assert(!face.lost);

        face.match_database_index = element_id;
    }
}

Run the project. The recognition results will be displayed in the console. If a face is recognized, you'll see the face id and name of a recognized person from the database. If a face isn't recognized, you'll see the message "Match not found".

Display the Preview of the Recognized Face from the Database#

Display the image and name of a person from the database next to the face in a video stream. In drawfunction.h, add a reference to the database, because we will need it when rendering the recognition results.

drawfunction.h

#include "database.h"

class DrawFunction
{
public:
    DrawFunction();

    static QImage Draw(
        const Worker::DrawingData &data,
        const Database &database);
};

In drawfunction.cpp modify the DrawFunction::Draw function by passing the database to it.

drawfunction.cpp

// static
QImage DrawFunction::Draw(
    const Worker::DrawingData &data,
    const Database &database)
{
    ...
    const pbio::RawSample& sample = *face.sample;
    QPen pen;
}

Save the bounding rectangle in the pbio::RawSample::Rectangle structure. Pass its parameters (x, y, width, height) to the QRect rect object.

drawfunction.cpp

QImage DrawFunction::Draw(...)
{
    ...
    // Save the face bounding rectangle
    const pbio::RawSample::Rectangle bounding_box = sample.getRectangle();
    QRect rect(bounding_box.x, bounding_box.y, bounding_box.width, bounding_box.height);
}

Create a boolean variable recognized that indicates whether a face is recognized or unrecognized. If a face is recognized, the bounding rectangle is green, otherwise it is red.

drawfunction.cpp

QImage DrawFunction::Draw(...)
{
    ...
    const bool recognized = face.match_database_index >= 0;

    const QColor color = recognized ?
        Qt::green :
        Qt::red;  // Unrecognized faces are highlighted with red

    // Display the face bounding rectangle
    {
        pen.setWidth(3);
        pen.setColor(color);
        painter.setPen(pen);
        painter.drawRect(rect);
    }
}

Get a relevant image from the database for a preview by face.match_database_index. Calculate the position of a preview in the frame.

drawfunction.cpp

QImage DrawFunction::Draw(...)
{
    ...
        // Display the image from the database
        if (recognized)
        {
            const QImage thumbnail = database.thumbnails[face.match_database_index];

            // Calculate the preview position
            QPoint preview_pos(
                rect.x() + rect.width() + pen.width(),
                rect.top());
}

Draw an image from the database in the preview. Create the QImage face_preview object that is higher than thumbnail on text_bar_height. The original preview image is drawn in the (0, 0) position. As a result, we get a preview with a black rectangle at the bottom with the name of a person. Set the font parameters, calculate the position of a text and display the text in the preview.

drawfunction.cpp

QImage DrawFunction::Draw(...)
{
    ...
    // Display the image from the database
    if (recognized)
    {
        ...
        const int text_bar_height = 20;

        QImage face_preview(
            QSize(thumbnail.width(), thumbnail.height() + text_bar_height),
            QImage::Format_RGB888);
        face_preview.fill(Qt::black);

        {
            const int font_size = 14;

            QPainter painter_preview(&face_preview);

            painter_preview.drawImage(QPoint(0, 0), thumbnail);

            painter_preview.setFont(QFont("Arial", font_size, QFont::Medium));
            pen.setColor(Qt::white);
            painter_preview.setPen(pen);
            painter_preview.drawText(
            QPoint(0, thumbnail.height() + text_bar_height - (text_bar_height - font_size) / 2),
            database.names[face.match_database_index]);
        }
    }
}

Draw face_preview in the frame using the drawPixmap method.

drawfunction.cpp

// static
QImage DrawFunction::Draw(...)
{
    ...
 
    // Display the image from the database
    if (recognized)
    {
        ...
        QPixmap pixmap;
        pixmap.convertFromImage(face_preview);

        painter.drawPixmap(preview_pos, pixmap);
    }
}

In worker.h, add a method that returns the reference to the database.

worker.h

class Worker
{
public:
    ...

    void getDataToDraw(DrawingData& data);

    const Database& getDatabase() const
    {
        return _database;
    }
};

Modify the call to DrawFunction::Draw by passing the database to it.

viewwindow.cpp

void ViewWindow::draw()
{
    ...
    const QImage image = DrawFunction::Draw(data, _worker->getDatabase());

    ui->frame->setPixmap(QPixmap::fromImage(image));
}

Run the project. If a face is recognized, it will be highlighted with a green rectangle, and you'll see a preview of an image from the database and a person's name. Unrecognized faces will be highlighted with a red rectangle.