Original article can be found here (source): Artificial Intelligence on Medium
Microsoft is not coming slow towards computer vision face API.
This article is not just suitable for the know-how of what type of work Microsoft did in the past few years for face Api but also give an intuition about different concepts of face detection and recognition.
After reading this you may be able to invent your own idea regarding the face application.
According to the annual report of “Stanford institute of Human-centered artificial intelligence,” the facial recognition received the 6.0 % share of global investment last year which is equal to 4.7 billion dollars in amount.
Now you can estimate the importance of facial recognition in the modern age. Either it can be in Large store of food, Security departments for surveillance, Home to allow only authorized persons etc.
Characteristics of facial data:
Locations: Location of the face means the actual position of a face based on the formation of pixels.
Landmarks: Facial landmarks considered as a detailed explanation of face. API of the face can be able to identify 27 landmarks on the face including nose, eyes, eyebrows etc.
Attributes: Attributes of the face are the predefined characteristics of face such as age, gender, facial hair, emotion etc.
Let’s come towards the intended topic.
Face API in Azure consists of 5 categories:
(a) Verification of the face.
(b) Finding similarities between faces.
(c) Identification of the face.
(d) Detection of the face.
(e) Clustering of the face based on visual similarities.
Now we discuss how all these differences with each other.
Facial detection assists to manifest the location of a face by using bounding boxes having X and Y coordinates. Face detection determines the position of a face in the image or video.
Face list is basically a list consists of a group of faces. Manage lists of faces that are similar to each other. You may want to find a face similar in the list of faces of family members, celebrities or friends etc.
Face list helps when you are working on face similarity and face identification.
Face verification can be used to check either detected face fulfil the predefined criteria. In other words, we can say do 2 images of a person’s face belongs to the same person or not. For example, such kind of approach can be used to enter in fully surveillance system that verified the face of just authorized persons.
Face identification vs Face similarity:
Face identification is used to identify a person by detecting face, either person information is available in the database of people or not.
Different face image of a person can be used for identification purpose against the person group let say “my inner circle” to check either test face image match in the person group or not.
Facial similarity means that 2 faces have similar attributes but not identical to each other.
Two working modes of face API: match person and match face.
· Match face will return similar faces after neglecting the same person threshold.
· Match person will return faces that are similar after applying the same person threshold.
Face morphing define as the transition of face in an image to another face. Warping both images and gradually transfer control points in first image location to location in second.
Note: Face morphing is not included in the Microsoft Azure face API. It is given for understanding.
We’ve covered quite a bit of information about the human face. You can use the testing web interface to learn how the calls to the API work. Also, consider using a client application to work through the concepts. Try this C# tutorial with a sample starter application.
If you want to dive deeper or find more information, the references are given below that you may find helpful for further understanding.