What is the working principle of face recognition system?

In today's society, face recognition systems are everywhere. Whether it's access control in and out of the office building or taking the subway, you can brush your face and take the bus. Face recognition system greatly improves the efficiency of traffic and is a very advanced technology. The public has always been curious about the working principle of the face recognition system and thinks it is a black technology. But in fact, seriously speaking, he is just a probability problem of mathematical operation. The working principle of face recognition system is mainly composed of the following parts.

First, the deep learning model.

The core and soul of face recognition system is the neural network model of deep learning. The so-called neural network model is actually an arithmetic unit. In this arithmetic unit, we can regard it as a black box, in which many parameters are stored, and these parameters can be adjusted automatically. This learning mode is mainly used for training. The purpose of training is to realize that the probability of the output result is infinitely close to 1 after inputting two photos of a person.

Second, the model training process.

The purpose of training the learning model is to make him remember the features of faces. The usual practice is to use a lot of face data, put these marked data into this model, and then tell it which person's photo is the same as another person's. Through constant training, he remembered the characteristics of people. Shows the parameters in the learning model, which are ultimately fixed. When we input two untrained pictures, it can also calculate the probability that the two pictures are a person.

Third, self-renewal system.

Another important part of the face recognition system is that it can update itself and learn by itself. When he judges for the first time 1 that two photos are the same person, he will use these two photos as his training set and update the parameters in his model, which is equivalent to remembering this person. If this man comes back later, he will be tried soon.

Face recognition system is the epitome of the development of deep learning and computer science in recent years, and its principle is very complicated.