Did you know that genre and race are some human characteristics that can crash the softwares of the face recognition? Let’s see how and why.
What is the face recognition?
It is a technic which was born for the safety of the structures and softwares which is now becoming ever more spreaded.
It’s ever more spreaded everywhere – from the shopping centers to the parks – it has its pros and cons, obviusly, especially if we use it for commerical or marketing purposes.
The face recognition is one of the technological protagonist of our time and it’s focused on a biometric technic to identify uniquely a person by checking and analyzing the models of the “face comparision”.
There are many technics. For example there is the “generalized” face recognition and the “regional adaptive” one. Basically it’s focused on the nodal points of the human face.
The values measured according to the points of the face of a person help to indentify or to verify a person uniquevoly. With this technic, the applications can help the datas obtained by tha face and they can carefully and quickly identify the person.
Advantages and disadvantages
Between the advantages of this technology there is the fact that it’s a kind of system “touch free”. The images of the face can be catched from distance and they can be analyzed without any interaction with the person.
The face recognition can be a great safety measure for the detection of the time and of the partecipation of a person to an avent or a place. Furthermore, it is even a cheap technology, because the elaboration is cheaper than other biometric technics.
But it’s not all sunshine. Infact, between the disadvantages of this technic there is for example the fact that the people can be identify only with perfect light conditions.
But that’s not all. Often the face recognition doesn’t recognise the face expressions, so to use we have to have always the same face expression, which is unlikely.
How does the face recognition work
Like we said there are many kinds of face recognition. The basic one, that we find for example on the smartphones, looks for the characteristics that define the face (eyes, nose, mouth) and it computes in a research to reproduce these aspects in a general combo of numbers.
It will so create a specific algorythm that will hang the selected face and that will autonomously recognise it.
By following this basic system, some devices have an internal device which will measure the distance between the points of the face by taking a picture of the face of the owner of the device. Through the shot the face will be saved and then it will use this data to unlock the device.
Is the face recognition precise?
Unfortunatelly not yet. Even though the progresses of the last years, the algorythms of automatically learning aren’t so precise yet and, for example, they can discriminate someone for genre or race or they can have some difficulty to recognise transgender people.
The algorythms of face recognition follow some models of evaluation and/or interpretation made thanks to the elaboration of big quantity of datas.
For this reason it is essential to teach the machines by using as much datas as possible with every human shadow. When this doesn’t happen, and they exclude some categories of people, the face recognition isn’t very precise.
But not only. Often the devices for the face recognition and the selection of the person through the face aren’t able to distinguish between man and woman. Let’s imagine what can happen with a transgender person.
Conclusions
Basically, even if it is useful to talk about the most common coding errors of the face recognition, it’s important to clarify the countless difficulties to code a machine to recognise the deversity and the thousands shadows of the human body.
The devices of face recognition are surely very useful and the technology is doing giant steps about it. Like often happens though the human being has to face with the machines that don’t have a soul and they don’t think about the whole human genre, otherwise they generalize it by making it a standard and not unique.
This post is also available in: