Weekly Update: 10-05-2021

Last Week we saw various Face detection Models,But the important thing is to recognize those detected Faces. There are various Face Recognition Models out there. After Various studies, We have Chosen Dlib's Face Recognition model for our Project. 

Contribution:

Tharun K -  Working of Euclidean Distance

Prashanth K - Open source Face_Recognition/deepface  models

Manjunath A - Code Implementation

Hanumesh V T - Evaluating Test Cases. 

How do we differentiate between Two Faces?

Every Face has its own unique descriptors called Face Descriptors. The Above Image Shows Face Descriptors and their respective Values. These Values are Unique for every Face . 

In Dlib's model, There are 128 Face Descriptors. These Descriptors are stored in a 128-Bit Array .So When Two Faces are to be compared ,It is Done using Euclidean Distance.

Fig. Formula to Measure Euclidean Distance.

The Euclidean Distance tells us How Close or Similiar two given faces are. The Lower the distance between them ,the similiar the Faces are.The Figure gives an idea on how the Learning takes place.

EXAMPLES:

Below are some of the examples of Facial Recognition .

The Faces in the Database are compared to the test Face .It shows the name if it recognizes the face else shows Unknown.

Code Snippets:

In the above Code , We load our data and store the known faces and their labels. The Faces are encoded into an 128-dimension and appended to the Known_face_encodings List. This will then be used for quick comparsion,which is shown below.

A lot of pre-processing happen in the above code for efficient Recognition, But the major part is the Comparison of Faces. The Test Image and its encoding is compared to each face in the known_face_encoding  using the Euclidean Distance . 

The minimum distance provides us the best match and hence its name is output and the face is thus Recognized and Labelled. The outputs were stored and discussed as shown previously.