Face-Recognition

What is Face-Recognition?

In this post, we are not talking about face recognition theory. Instead, we are going to talk about the module face-recognition.
It works as same as face recognition. Let’s see how to recognize face by the module.

Steps of Face-Recognition

---

1. Load images and learn how to recognize it. The images can be loaded by face_recognition.load_image_file(file) where:
   

   • file : image file name or file object to load.
     
     

2. Convert the images to prepare comparsion group. The images can be converted by face_recognition.face_encodings(face_image, known_face_locations=None, num_jitters=1, model=’small’) where:
   

   • face_image : Input image.

   • known_face_locations : The bounding boxes of each face if you already know them.

   • num_jilters : How many times to re-sample the face when calculating encoding.

   • model : Which model to use. “Large” or “Small”
     
     

3. Create two lits of known face encodings and their names.

4. Get the face location of the target image by face_recognition.face_locations(img, number_of_times_to_upsample=1, model=’hog’) where:
   

   • img : Input iamge.

   • number_of_times_to_upsample : How many times to upsample the image looking for faces.

   • model : Which face detection model to use. ‘hog’ is less accurate but faster on CPUs. ‘cnn’ is a more accurate deep-learning model which is GPU/CUDA accelerated.
     
     

5. Encoding the image by usingface_recognition.face_encodings() again. Note that we should use the tuples that obtain from Step 4.

6. Compare the data between we obtained at Step 5 and we prepared at Step 3 by using face_recognition.compare_faces(known_face_encodings, face_encoding_to_check, tolerance=0.6) where:
   

   • known_face_encodings : A list of known face encodings.

   • face_encoding_to_check : A single face encoding to compare against the list.

   • tolerance : How much distance between faces to consider it a match.

Implementation

• Face Recognition