The Knowledge of Face Recognition

Facial recognition technology has developed rapidly in just a few years of the 21st

century. According to the research and test of the National Institute of Standards

and Technology (NIST), the deadline is April 2020. The error rate of the facial

recognition algorithm It is 0.08%. This error rate is when this algorithm is the best

face recognition algorithm. Under the same conditions, the facial recognition algorithm

in 2014 was only 4.1%. Enough to see the development speed of facial recognition.

What is biometric facial recognition?

Face recognition technology is based on the facial features of a person, and the input

face image or video stream is first judged whether there is a face, if there is a face,

then the position, size and main points of each face are further given. Location information

of facial organs. Based on this information, the identity features contained in each face are

further extracted and compared with known faces to identify the identity of each face.

The broad sense of face recognition android actually includes a series of related technologies

for building a face recognition system, including face image collection, face positioning,

face recognition preprocessing, identity confirmation, and identity search, etc.; while the

narrow sense of face recognition machine specifically refers to the adoption of A

technology or system for identifying or searching for human faces.

The biological characteristics studied by biometric recognition technology include face,

fingerprint, palm print, iris, retina, voice (voice), body shape, personal habits (such as

the strength and frequency of typing on the keyboard, signature), etc. The corresponding

recognition technology has people Face recognition, fingerprint recognition, palmprint

recognition, iris recognition, retina recognition, voice recognition (voice recognition can

be used for identity recognition and voice content recognition, only the former belongs

to biometric recognition technology), body shape recognition, keyboard strokes

Identification, signature recognition, etc.

Best Face recognition software is to compare the facial features to be

recognized with the obtained facial feature template, and judge the identity

information of the face based on the degree of similarity. This process is

divided into two categories: one is confirmation, which is a process of one-to-one

image comparison, and the other is identification, which is a process of one-to-many

image matching and comparison.

Face image collection and detection

Face image collection: Different face images can be collected through the

camera lens, such as static images, dynamic images, different positions,

different expressions, etc. can be well collected. When the user is within

the shooting range of the capture device, the capture device will automatically

search for and shoot the user's face image.

Face detection: In practice, face detection is mainly used for preprocessing

of AI face recognition, that is, to accurately calibrate the position and size

of the face in the image. The pattern features contained in face images are

very rich, such as histogram features, color features, template features,

structural features, and Haar features. Face detection is to pick out the

useful information, and use these features to realize face detection.

How accurate is facial recognition?

The visual artificial intelligence industry is driven by technology, and the

core of the technology lies in the three aspects of data, computing power,

and algorithms. The emergence of GPU and AI dedicated chips has broken

through the computing power bottleneck of traditional CPUs, and data

computing speed and processing scale have exploded. Big data analysis

provides hardware support. More and more application fields are accumulating

increasingly rich big data. Massive image and video content provide powerful

data support for deep learning.

The emergence of deep learning has greatly promoted the development of the

visual artificial intelligence industry. In 2015, in the Imagenet competition

of the visual artificial intelligence system recognition project, Resnet surpassed

the 5.19% of human vision for the first time with a recognition error rate of

3.57%. At present, the accuracy of 3d face recognition has been improved to

over 97%.

At present, the application of visual artificial intelligence on a global scale is

concentrated in the fields of smart consumption and smart manufacturing, with

remarkable results and the continuous expansion of subdivisions. With the

continuous development of technology, visual artificial intelligence can recognize

the types of information from the initial text information to the recognition of

human faces, human postures, and various objects. The recognition accuracy has

also changed from the initial 1:1 comparison to the 1:N comparison used in access

control systems, etc., and the M:N dynamic monitoring used in blacklist monitoring

and other scenarios. At the same time, the automation of data labeling has been

greatly improved. Further improve identification efficiency and reduce identification

costs.

Due to the popularity of deep learning, machine learning research has flourished

in recent years, and real time face recognition technology has also been greatly

improved. In a typical use case, faces in photos, videos or real-time streaming media

are scanned and analyzed, and then their features are compared with the annotated

faces in the database.

How to ensure the accuracy of face recognition Device?


This technology is being used to combat human trafficking and rapid airport security

checks, while it is also being used to monitor concerts and sports events.

However, the accuracy of facial recognition is still an issue. Researchers are beginning

to worry about discrimination and prejudice in artificial intelligence systems. The

technology still has major flaws in correctly identifying people of color and women.

One of the reasons for this problem is the disparity in the proportion of men relative

to women and whites relative to colored people in the data set.

For machine training, data diversity is important, but the size of the data is also

important. The training and testing of face recognition systems need to be carried

out on millions or even tens of millions of faces.

For many years, researchers have been conducting related research through face

recognition security system data sets. This data set containing image links is organized

and generated from a resource pack. This resource pack is used for various scientific

project research, including research on estimating the location of photos and videos

without using geographic coordinates.

IBM has developed a new project called "Face Diversity" and prepared millions of

pictures for it. This project will further improve the fairness and accuracy of face

recognition.

Researchers collected and annotated photos of various objects through the Internet

to train computers to better understand the world around them.

Usually, they get a huge number of pictures through Google image search and some

other ways. The resulting data set is usually used for academic research, such as

training or testing face recognition algorithms. But with companies such as

Microsoft, Amazon, Facebook and Google betting on artificial intelligence, facial

recognition is moving out of the laboratory and entering the field of vision of

large companies.

As consumers realize the immense power of the data they leave on the Internet,

face recognition data sets are exacerbating concerns about privacy and surveillance.

Therefore, some researchers are re-examining this brutal collection of photos of

others. In the Internet full of sharing spirit, the use of other people's photos

should have their consent.

Many companies, research institutions, and individuals have compiled data sets for

facial recognition, and IBM is just one of them. Some of these data sets are

composed of actual images, and there are some IBM-like data sets that are

composed of image links. Sometimes, the data set can also be obtained by shooting

models.


Normally, these facial recognitiondata sets are knowledge-sharing, but they must

be used for non-commercial purposes, such as algorithm research. But a large number

of similar data sets can be downloaded for free from the website.

The facial recognition knowledge sharing agreement was first released in 2002,

far before the current artificial intelligence boom.

Although the researchers use pictures for free on the website, they also

admit that many people who upload these pictures may be surprised by

the fact that the pictures are used to train artificial intelligence.

Some china facial recognition researchers believe that people should decide

for themselves  whether images can be used for computer vision or artificial

intelligence research through authorization.

In this regard, the face recognitionknowledge sharing agreement does not

help much. As long as the relevant terms are followed, this license agreement

from a non-profit organization does not restrict any form of artificial intelligence

development.

The CEO of Creative Commons stated: "These agreements are not designed

to protect privacy or research ethics."

In recent years, artificial face recognition intelligence has developed so fast

that the relevant regulations have hardly been formulated, let alone implemented.

Legally, the company has no obligation to inform when collecting and using images for

facial recognition.

No relevant facial recognitiion federal regulations have been issued yet. In each state, the

situation is different: For example, Illinois has a law that requires companies

to obtain customer consent before collecting biometric information; the state

Senate in Washington, where Amazon and Microsoft are headquartered, recently

passed a restriction on facial expressions. To identify the use of the bill, the

bill still needs to be passed by the State House of Representatives.

In March of this year, the Senate introduced a bill requiring companies to

obtain consumer consent before collecting and sharing identification data. It

also requires the company to conduct external tests to ensure that the

algorithm is fair before implementation.

The facial recognition technical policy director of the Electronic Frontier

Foundation, a digital copyright organization, said that even if there are no strict

laws restricting the use of private photos for artificial intelligence training, companies and

research groups should pay attention to ethics.

In his view, Facial Recognition device this means that the use of photos requires the explicit consent

of the people in the photos. Even if this is difficult to do, it is a reality that

companies must face.

website: www.china-attendance.com

website: www.hfteco.com

Youtube: Huifan Technology

Linkedin: Huifan Technology

Facebook: Huifan Technology

Email: info@hfcctv.com