anhminh3105 / FaceNet_FaceRecognition

FaceNet_FaceRecognition

Real Time Face Recognition by FaceNet

This application used the FaceNet library of @davidsandberg to implement the real time face recognition on the your computer's camera. This implementation made several changes to face.py and real_time_face_recognition.py. to adapt to this modification.

The general procedures are:

1. Detect faces and their landmarks in video input stream by MTCNN.
2. Align detected face images.
3. Contruct embeddings for each of these face images.
4. Construct an embedding database of people you want to recognise.
5. Compare these face embeddings with your embedding database and find the matching person.

To run this application:

0. Download the pre-trained model 20180402-114759 to folder model_checkpoints/
1. Make a folder of images of people you want to recognise (one person for each image) and name the images with respect to the corresponding people.
2. Clone the repository to your local computer.
3. Install the library and framework requirements and set up the PYTHONPATH variable to the src/ folder.
4. In distributed/real_time_face_recognition.py, provide the full path of the image folder you created earlider to variable img_folder
5. Run the python file with command C:\...distributed> python real_time_face_recognition.py

Results: