QFaceRecognition

Qt/Qml lib for face detection/recogntion based on dlib

Features

• command line base recogntion
• works with image files directly
• video filter for recognize face works whit Qt Camera
• Qml support

Build And Run

clone

clone project with it's submdoule "dlib" using following bash code

git clone --recurse-submodules https://github.com/mshobeyri/QFaceRecognition.git

build & run

first get dlib and build it using dlib build manual form here. but if you use qtcreator you can just open CMakeLists.txt file in dlib and build dlib.

then copy dlib.lib file from "dlib build folder" in "lib" folder in root of project.

then you can run project.

usage

remember to install git lfs to get model files and set their pathes to QFaceRecognition class using setModel function. something like this:

QFaceRecognition::setModel(
        "../../model/dlib_face_recognition_resnet_model_v1.dat",
        "../../model/shape_predictor_5_face_landmarks.dat");

Examples

console face recognition

the code api is so simple:

int
main(int argc, char* argv[]) {
     QFaceRecognition f;
    f.introduceFolder(argv[1]); //known images folder path
    qDebug() << f.recognizeFolder(argv[2]);  unknown images folder path
	}

you can run the project like this:

./ConsoleFaceRecognition <known image folder> <unknown images folder>

example:

./ConsoleFaceRecognition ../../assets/known ../../assets/unknown

note: set the pathes relative.

the output is name of known files paired with path of unknown files.

qml face recognition

this is a real time face recognition project whit an output like this:

you can see using video filter example in qml code here

import QtQuick 2.0
import QtQuick.Window 2.0
import FaceRecognition 1.0
import QtMultimedia 5.2

Window {
    id: iwin
    visible: true
    width: 640
    height: 480
    title: qsTr("Face Recognition")    
	Camera
    {
        id:camera
    }
    FaceRecognitionFilter{
        id: frFilter

        captureRect: {
            videoOutput.contentRect;
            videoOutput.sourceRect;
            return videoOutput.mapRectToSource(
                        videoOutput.mapNormalizedRectToItem(Qt.rect(
                                                                0, 0, 1, 1
                                                                )));
        }
        faceRecognition{
            mode: FaceRecognition.Recognize
            distanceThreshold: 0.6
            onRecognizeProcessEnded: {
                console.log(names, positions)
            }
        }
    }

    VideoOutput{
        id: videoOutput

        source: camera
        autoOrientation: true
        fillMode: VideoOutput.Stretch
        filters: [ frFilter ]
    }
}

---

Note:

• use mode to switch ability of application form Detecting and Recognizing.
• use distanceThreshold to change threshold of accepting two face is from same person.
• you can limit capture image to captureRect section

these are the signals you can use:

signals:
    void faceDetected(QRect position);
    void faceRecognized(QString name, QRect position);
    void detectionProcessEnded(QList<QRect> positions); //emit just when mode is detection
    void recognizeProcessEnded(QStringList names, QStringList positions);

---

thanks

special thanks to saeed khodaygani