Face-Recognition-using-FaceNet

This face recognition system is implemented upon a pre-trained FaceNet model achieving a state-of-the-art accuracy. The system comes with both Live recognition & Image recognition. It is trained on faces of some celebrities.

For any queries Contact: Ankur Goswami

• Installing dependencies:

  • For Anaconda users: conda install --file requirements.txt
    
  • For python users: pip install -r requirements.txt
    (even Anaconda users can use this if they use anaconda prompt instead of terminal)

• Downloading the model:
  The repository requires an additional file to work. The file is too large to upload here. So I've provided a Google Drive link of it. Download the file and keep it inside /data/model/ directory.
  Click Here to download the file.

• Training on other faces:
  To train model on different faces, follow the given steps:
  

  1. Put the images containing clear frontal face in /data/images/ directory.
  2. Open the repository directory in terminal and run following commands in given order:
     1. cd script
     2. python generate_data.py
  3. Follow program instructions.

• Testing/Detecting faces:
  

  1. Face Recognition from Images:

     1. Put the images containing the faces to predict in /test/ directory.
     2. Open the repository directory in terminal and run following command:

         python image_recognition.py
     

     1. Output images will then be available in /test/predicted/ directory.

  2. Live Face Recognition(Obviously using camera):
     Open the repository directory in terminal and run following command:

     python live_recognition.py
     

Examples:

NOTE: Faces with Unidentified labels are faces on which the model is not trained.

Example #1:
Before:

After:

Example #2:
Before:

After:

Example #3:
Before:

After:

Example #4:
Before:

After:(Need to zoom)

Example #5:
In this example, the model was trained on faces of my friends.
Before:

After: