Facial recognition: top 7 trends (tech, vendors, use cases)

In recent years, the appearance has been greatly concerned and surprising.

But later on this.

In this web document, we ask about seven precedents and trends to identify people who create 2021 landscapes.

1. Best technology and supplier
2. The impact of AI is larger than others
3. 2019-2024 market and main application scenarios
4. Face recognition in China, India, the United States, EU, British, Brazil and Russia
5. Confidential vs. security: Is it free, freezing, regulation or prohibition?
6. The latest hacking: Is it possible to deceive the face recognition system?
7. Toward a hybrid conclusion

Let's enter the lesson at once.

How facial recognition works

Face recognition is a process of identifying or testing the person's identity based on the face. Amplify, analyze, and associate the pattern based on the details of the human face.

1. The face detection process is an important step to determine the position by detecting images and human faces in the video.
2. The face capture process converts analog information (face) into digital information (data or vector) based on a person's face.
3. In the face comparison process, check if two faces belong to one or the same person.

Let's explain these three stages in recent cases.

A student with Washington identifies more than 6, 000 facial images from 827 videos posted on the website parler during the event that occurred on January 6th, using the Face Mining app on January 6. (Source: WIRED 20 JANUARY 2021).

1. Protests, riot members, and correspondents contributed to the actions of shooting individuals using their own mobile phones (analog identity became a digital image).
2. Extract the face from 200. 000 images using face detection.
3. FBI wants to investigate, portrait (conversion from digital pixels to vector), and very well, compare and identify faces with existing databases (supported by the AFIS/ABIS system).

Today, this is the naturalest in all biometrics measurements.

For example, you recognize yourself with your face, not fingerprint or iris of your eyes.

Before going ahead, let's simply define the two major sentences, "identification" and "authentication".

Face recognition data to identify and verify

Biometrics is used to identify and authenticate that person using its own known verification data.

For a definition of biometrics, please read the web document on biometrics.

Identification is answering the question: "Who are you?"

Authentication is answering the question: "Who are you?"

Stay tuned. Here are some examples:

• In facial biometrics, a 2D or 3D sensor "captures" the face. It is then converted into digital data using algorithms and the resulting image is compared to one stored in a database.
• These automated systems can be used to identify or verify a person's identity in just a few seconds based on their facial features (geometry). They can do this in crowds and in dynamic, volatile environments.

• iPhone X owners are already familiar with facial recognition technology.

Of course, other signatures of the human body are included, such as fingerprints, iris scans, voice recognition, digitization of palm veins, and behavioral measurements.

Why face recognition, then?

Facial biometrics remain the preferred biometric standard.

This is because it is easy to implement and deploy. There is no physical interaction with the end user.

Moreover, the process of face recognition and matching/identification is very fast.

So what is the best face recognition software?

#1 Top facial recognition technologies

In the race of biometric innovation, several projects are vying for supremacy.

Google, Apple, Facebook, Amazon and Microsoft (GAFAM) are also active in this race.

All software giants regularly publish theoretical discoveries in the field of artificial intelligence, pattern recognition and face analysis, deepening our understanding as quickly as possible.

Let's take a closer look:

Academia

The GaussianFace algorithm, developed in 2014 by researchers at the Chinese University of Hong Kong, achieved a success rate of 98, 52% in face recognition (compared to 97, 53% for humans). It's an impressive score, despite the drawbacks of memory requirements and computation time.

Facebook and Google

In 2014, Facebook introduced DeepFace, which can determine with an accuracy of 97, 25% whether two photographed faces are the same person. If a human takes the same test, they will get it right 97, 53% of the time, which is only 0, 28% better than Facebook's program.

In June 2015, Google Facenet was improved. On an expanded dataset, the Labeled Faces in the Wild (LFW) facenet achieved a new record accuracy of 99, 63% (0, 9963 ± 0, 0009). Using artificial origins, neural networks and new methods, the company Mountain-Voy was able to connect personalities to their owners with literally perfect results.

This development was integrated into Google Photos and is used to classify and self-create photos based on recognized people.

The company, having proven itself important in the field of biometrics, soon released an unofficial version of OpenFace with unheated initial code.

Microsoft, IBM, and Megvii

A study conducted by scientists from the Massachusetts Institute of Technology in February 2018 showed that Microsoft, IBM and the Chinese company Megvii (Face++) showed an increased rate of pluckers when identifying dark-skinned women to compare with honest pinsants.

At the end of June 2018, Microsoft announced on its blog that it had significantly improved its own technology for determining individuals with biased attitudes.

Amazon

In May 2018, ARS Technica stated that Amazon is actually promoting a personal cloud service for determining people under recognition among law enforcement agencies. The conclusion is that it can recognize up to 100 people in a single image and compare people with a database of 10 million people.

In July 2018, Newswee k-newspaper stated that Amazon's development of the definition incorrectly recognized 28 members of the US Congress as people who had been arrested for crimes.

Key biometric matching technology providers

In late May 2018, the US Department of Science and National Security published the results of sponsored tests at the Maryland Testing Center (MDTF). The standard test measured the performance of 12 systems to determine people in a 2m x 2, 5m corridor.

The solution, which applied Thales's Person Identification Software (LFI), showed excellent results: the person identification coefficient was 99. 44% in the shortest 5 seconds, with an average index of 66%. The missing coefficient was 1% of the equation, with an average of 32%.

March 2018 - A technology was discovered to determine the best performance of an individual during role testing of more than 300 volunteers.

More about the performance sample: A report by NIST (National Institute of Stereotype Technology), established in November 2018, details the accuracy of identification of 127 algorithms and member names.

The auxiliary results are emphasized by the current analysis of the NIST face recognition vendor test (FRVT) 3 that took place in late 2019. See the NIST report.

NIST also indicated that the best way to judge individuals has no race or sexual bias, and ITIF declared this in January 2020. The critic made a mistake

The report of NIST (August 2020 and March 2021) is a report entitled "The accuracy of a person judgment by a mask that supports the algorithms created later by the COVID-19", and how the method is. It indicates whether to increase your productivity.

Facial Emotion Recognition (FER)

The decision on the face impression by rea l-time or still images is to support the support of image processing software and identify these impressions as an impression of the support, joy, anger, surprise, shock and sadness, or these combinations. It is a process that reflects the facial expression of the face to do.

There are three steps in recognition and interpretation of human impressions:

• 1) Face detection
• 2) Determination of expression
• 3) Assign the expression to a specific sensual state.

The popularity of determining the impression of the face is justified by a wide range of areas that are likely to be used.

This is different from the definition of a person aimed at identifying a person, not an impression.

Face expressions can be expressed in geometric or external symptoms, parameters extracted from reborn images, personal faces, dynamic models, 3D models, etc.

Vendors include KAIROS (recognition of face and impression for brand marketing), noldus, IMPACTIVA, and SightCorp.

For more information about face impression recognition (Fer), see the EU technical document in May 2021.

#2 Learning to learn through deep learning

The general line of all data through revolutionary technology is a fake mind (AI) or rather a thorough study if the system has the ability to study on a database.

Why is that important?

This is the center of the latest algorithm created by Tales and other major companies. This hides a secret that determines a person, tracks a person, compares people, and translates conversations in real time.

All personal judgment systems are made from different things.

According to NIST's recent reports, the definition accuracy has been significantly improved in the past five years (2013 to 2018), exceeding the features of 2010 to 2013.

Most of the algorithms that judge individuals in 2018 are ahead of the most correct method at the end of 2013.

In fact, during the 2018 test, NIST pointed out that when searching for data from 26 million photos, it was not possible to find the correct image in a case of 0, 2%, but in 2014 this indicator was 4. It was %.

A 2020 NIST study found that the best facial identification methods had a 0. 08% error rate, or less than one error per 1, 000 images. (Source: Accuracy of Facial Identification Systems, CSIS)

So, did you know this?

That's a 50-fold improvement in six years.

Think about it this way:

Artificial neural network algorithms can make facial detection methods more accurate.

#3 Facial recognition markets

Face recognition markets

A June 2019 study estimated that the global ID bazaar will generate revenue of US$7 billion from 2019 to 2024, at a compound annual growth rate (CAGR) of 16%.

The market size was estimated at $3. 2 billion in 2019.

The two main reasons for the increase are seen as surveillance in the municipal sector and a host of other applications in various parts of the market.

According to the survey, the main vendors of facial recognition systems are:

Accenture, Aware, BioID, Certibio, Fujitsu, Fulcrum Biometrics, Thales, HYPR, Idemia, Leidos, M2SYS, NEC, Nuance, Phonexia, Smilepass.

The main application areas of facial detection systems are divided into three.

What is facial recognition used for?

Here are three main categories of applications that use facial detection.

1. Security - law enforcement

Forensic scientists are well-suited to match multiple types of biometric data using an Automated Biometric Identification System (ABIS).

This bazaar is justified by the increasing energy in the fight against crime and terrorism.

The advantage of facial recognition systems for the police is clear: the detection and prevention of atrocities.

• Facial recognition is used in issuing personal identification documents, usually in combination with other biometric technologies such as fingerprints (to prevent identity fraud and identity theft).
• Facial matching is used in border control to match the digitized biometric passport portrait with the holder's face. In 2017, Thales delivered new automated checkpoints for the PARAFE (Automated Fast Track Crossing at External Borders) system at Paris' Roissy-Charles de Gaulle airport. The agreement was signed to facilitate the transition from fingerprint detection to facial recognition in 2018.
• Facial biometrics are still sometimes used for police checks, but are tightly controlled in Europe. In 2016, the "man in the hat" suspect in the Brussels terrorist attacks was identified thanks to the FBI's facial recognition software. South Wales Police used it at the end of the UEFA Champions League in 2017.
• In the United States, 26 states (possibly 30) allow law enforcement to search driver's license and photo ID databases. The FBI has access to driver's license photos in 18 states.
• The combination of drones and aerial cameras offers an exciting combination for widespread use during public events: facial recognition. According to the Keesing Journal of Documents and Identity in June 2018, some hovering drone systems can carry a 10-kilogram camera lens that can identify suspects from a distance of 800 meters and up to a height of 100 meters. A power cable can connect the drone to the ground indefinitely. Communications with the ground are also line-based and cannot be intercepted.
• CCTV systems with facial recognition can improve the efficiency of public safety missions. Let's illustrate this with four examples:

1. Searching for missing children and disoriented adults.
2. Identifying and searching for exploited children.
3. {RUMMARY}Identifying and tracking criminals.
4. {ROUND}

1. Find Missing children and disoriented adults.

Aiding and expediting investigations.

CCTV systems with facial recognition can significantly speed up the work of operators by adding reference photos of the parents of a missing child and matching them with past appearances of that person seen in the video. Police can use facial recognition to search video sequences (called video analytics) that show the location and time when a child was reported missing.

Read how Delhi Police used facial recognition to track 3, 000 missing children in four days.

2. Identify and find exploited children.

Police can get a better understanding of how a child was moving before going missing and find where they were last seen. Police can check the accuracy of the recognition and make every effort to bring back the missing child. A similar process can also be applied to confused adults (e. g. those with dementia, memory loss, epilepsy, Alzheimer's disease).

Isolating a specific individual in a video sequence is crucial. This can also speed up the work of investigators in child exploitation cases.

3. Identify and track criminals.

Video analysis can help create timelines, track actions on maps, uncover details and uncover unobvious connections between participants in a case.

4. Support and accelerate investigations.

Video surveillance systems that have the definition of a person have the ability to apply it, and the police have the ability to track and identify criminals accused of aiding and abetting violation of the law. The police still have the opportunity to take preventive measures. By applying the image of a popular criminal from a video or external image (or database), the operator has every chance to identify and react to a coincidence in the actual video, it is not very late.

CCTV-systems with the definition of people have every chance to apply to the investigator to look for video evidence, and then compete.

2. Health

The ability to separate the accused and display the personal is very fundamental to speed up the investigation process by the investigator of the video evidence in the desired list of important details. They have every chance to take, different how the action was developed.

In this field, significant fortunes have been achieved.

• {space}.
• space
• {space}

3. Banking and retail

Maintaining anesthesia.

Undoubtedly, only in this field, the definition of a person has at least built hope. And yet, absolutely perhaps, it is the most promising.

The "KYC (Know Your Own Customer)" system that defines an individual in Internet mode will be a burning topic in 2021.

After all, in the United States, only in the second quarter of 2020, the first current account discovery was on the Internet (and 36% in the sector).

The pandemic accelerated this momentum, and for now, almost all sectors are covered.

In addition, the increase in the use of mobile devices inspires the company to focus on mobile first and develop all mobile that is beneficial to users of the user.

The selfie process calls for the development of a reliable determination of clarity to avoid gaps when introducing still images.

Liveliness detection effectively justifies that the manufactured selfie is of a living person.

To accommodate the current preferences of the buyer, financial institutions (F. I. S.) are investing in the digital registration of buyers through Internet and mobile channels.

The determination of a person with the ability to determine a living person simplifies the registration function and KYC in Internet mode. Thales is considered the most collaborative provider of personality test conclusions, with a wide range of quantities and features.

According to Forbes, the invention of the digital account (DAO) has become the most well-known technology in the banking sector in its third year. Nearly 80% of financial institutions will add new DAO systems or improve the ones they have in place in 2020 and 2021.

This important desire is mixed with the latest advertising achievements in customer proficiency.

By installing a camera at the store, it has become possible to analyze customer behavior and improve the purchase process.

Like the recently developed system, sellers can receive information on customers from profiles on social networks and make appropriate individual response based on them.

The US department store saxophone Fifth Avenue already uses such a system. It is also used in Amazon Go stores.

Is the appearance of a selfie board close?

Since 2017, King of Fly Chicken KFC in the United States and China's retail and Alibaba have been testing for personal recognition payment systems in Hangzhou (China).

In March 2021, the 52 Paths Stores (X5 Retail Group) started a fac e-t o-face no n-contact settlement of cash register terminals in cooperation with VISA and Severbank payment systems.

According to Yahoo, the payment system will be used at 3. 000 stores by the end of the year.

#4 Mapping of new users

According to Interfax Communications, Muscobian people will be able to pay by subway by the end of 2021.

Face recognition in China

Currently, the largest market in the human recognition function is in the United States, but in the Asi a-Pacific region is the fastest. In this field, China and India are leading.

From banks and airports to police, in China, human recognition technology has become a new topic.

Authorities are currently expanding programs that use fac e-recognized sunglasses, and police are starting to use sunglasses near Beijing.

China has also built and improved video monitoring networks nationwide.

According to CNBC, more than 200 million video surveillance cameras will be used in 2018, and more than 500 million units will be used in 2021.

The tower with facial recognition function in Chinese cities is a symbol of this movement.

This is due to the social lending system developed by the Chinese government.

In the top 10 cities with a large number of street cameras per capita, Spring, Shenzen, Shanghai, Tianjin, and Chiana are leading alone.

London is 6th and Atlanta is 10th, reported by Guardian on December 2, 2019.

According to the New York Times on April 14, 2019, the Chinese police are working with artificial intelligence companies such as YUTU, MEGVII, Sensetime, and Cloudwalk.

China's ambition in AI (and human recognition technology) is high. The country wants to be a world leader in A. I. by 2030.

Surprisingly, China has definitely protected personal biometric data and expand its government access to personal information.

Facial recognition in Asia

The phenomenon was proven by an expert on November 2, 2020 by an expert on the confidentiality of Emmanuel Pelene Reprei.

In the 2020 Tokyo Olympics (transfer in September 2021), personal authentication is essential.

This development identifies mechanically authenticated people, and can be accessed, enhancing their skills and security. In the sunshine, it is still used to simplify access to mobile banking.

In Sydney, the development of personal identification at the airport is progressing, and the flow of people can be greatly and more peaceful.

In India, the AADHAAR plan is considered the world's most comprehensive biometrics database. In the range of applications in March 2021, 129 billion people in the state already have an original digital identification number.

Widai, a serious organization, has announced that face authentication will be implemented in step.

In real time, it is tested for money (October 2020).

Face authentication is available at another moment of authentication, as a finger scale, or as an additional fusion service, such as a finger mark, a scale of the eyes.

In India, an extremely larg e-scale system to determine the people of the world will be expanded by 2021.

The National Farmers Registration Bureau (NCRB) has published a business request form to develop a nationwid e-scale personal judgment system.

According to a 16 0-page document, this system is a concentrated web application installed at the NCRB Center of Sepaya. Access to this system is not closed to all police sites.

Other large projects

This system mechanically identifies people from video recording and surveillance cameras. The bureau claims that the police will definitely help catch criminals, miss people, and confirm the corpse.

The Supreme Election Court (Tribunal Superior Eleitoral) has participated in Brazil's nationwide biometric data. The challenge of this plan is to create a biometric database and an original ID, and enter 140 million information there.

In Africa, Gabon, Cameroon, and Burkina Faso chose Tales to complete voters' identification work by biological authentication.

Since 2017, the Russian Central Bank has been developing programs nationwide to collect data on face, voting, rainbow shell scan, and fingerprint data.

However, the process is progressing quite slowly, according to information from the Biometrics update website on March 13, 2019.

By the end of 2019, the capital talked about one of the huge networks in the world of 160. 000 videos surveillance cameras, with technology to determine people to secure social security.

The implementation began in January 2020.

#5 When face recognition strengthens the legal system

Russian law does not regulate personal detection and analysis without consent.

Face recognition technology is fundamentally changing ethical and social issues over data protection.

Is such a technical development like science fiction really threaten our freedom?

E.U. and U.K. biometric data protection

Will it threaten our anonymity?

The General Data Protection Rules (GDPR) provides strict frameworks in Europe and the UK.

There is no doubt in any survey of citizens' private life and business transactions, and privacy infringement is a strict penalties.

The AVG, which will be implemented in May 2018, supports the principles of the European system that has been harmonized by protecting the right to be forgotten and aggressive discrimination.

That's right. This will apply one law to 500 million.

U.S. biometric data protection landscape

This has an international meaning.

While federal law is not enacted, cities and states are intervening to fill the gap.

Washington has become the third state (after Illinois and Texas), which formally protected biometric data under the new law passed in June 2017.

From January 2020, California has become the fourth state

The California Consumer Privacy Law (CCPA), which was passed in June 2018 and will come into effect on January 1, 2020, has a significant impact on California and nationwide privacy rights and consumer protection.

The law is often attracting attention as a model of the Federal Data Privacy Law.

In that sense, CCPA has the potential to be as important as AVG.

In July 2018, Microsoft's President Bradford L. Smith called for his research and use of his research and use, compared to a product like a strictly restricted drug.

In May 2019, Alexandria Ocaci o-Cortes Representative Cortes has expressed "absolute" concerns (the impact on our citizenship and freedom) at a committee hearing on face recognition technology.

The SHIELD (STOP HACKS and IMPROVE ELECTRONIC DATA SECURITY) law in New York has enforced on March 21, 2020. This law requires cyber security programs and protection measures for New Yorkers.

This law is applied to companies that collect personal information from New York people.

Facial recognition bans (San Francisco, Somerville, Oakland, San Diego, Boston, Portland)

Thanks to this law, New York has been on par with California.

As face authentication becomes an increasingly common method of law enforcement, concerns about privacy and citizenship are escalating in the United States. On May 6, 2019, San Francisco resolved to ban face authentication.

The resolution against surveillance signed by the San Francisco Board of Supervisors will prohibit city agencies, including the San Francisco Police Department, from using the technology after June 2019.

Yes, this applies to law enforcement agencies too.

According to the Boston Globe dated June 27, 2019, the Somerville (Massachusetts) City Council voted to ban personal recognition, becoming the second city to make such a decision.

• Space
• {space}
• {space}
• {space}
• {space}.
• space

The Virginia General Assembly passed a new bill (H. B. 2031) that prohibits law enforcement agencies from continuing to use software to recognize individuals after July 1, 2021.

Following the decisions of San Francisco, Somerville, Oakland, and now San Diego, Boston, and Portland, Devata has become a louder voice in many cities and states, not just in the United States.

How to better regulate emerging technologies?

• {Space}
• Space
• Space
• Space

Is there a political vacuum? At what level?

We will follow the outcome of all these discussions as the US Congress comes under pressure from activists calling for a ban on the technology and providers calling for regulation.

However, as of May 2021, there is still no federal legal framework to resolve this issue.

The European Union (EU) Committee is considering measures against face recognition technology indiscriminately. Ursla Paul Dare Lien, chairman of the European Commission, wants to coordinate the humanitarian and ethical results brought by artificial intelligence. She promised to announce a law on the origin of artificial intelligence in the near future.

The final version of the European Commission's document is available online. The European Commission proposed a strict regulation plan in April 2021. However, according to Reuters, it may take several years for the standard to come into effect.

Similarly, the EU's two supervisor (EDPB and EDPS) called for personal identification in public in June 2021.

Again, the problem of privacy, consent, and creep (the data collected for a certain purpose is used for another purpose) is the center of the debate.

India and its national biometric identification scheme, Aadhaar

For details on the Biometrics Data Protection Law (EU, the United Kingdom, the United States), you can see it in the Biometrics data file.

In India, the Supreme Court specified privacy rights in the National Constitution at the Puttaswami ruling on August 27, 2017. The conclusion has changed the balance between the citizens and the nation, and has raised a new issue to expand the planar plan.

However, on February 28, 2019, the Indian government confirmed the implementation of an EID biological certification program by personal organizations.

Recoil: Legal system and their professionals are even stronger.

#6 The rebels – facial recognition hackers

As ambassador and parents of data defense regulations, data defense workers have become indispensable for companies and are very required.

Despite these technical and legal weapons designed to protect data, people, and their anonymity, dangerous voices are still heard.

Some people are excited about this and sounds a warning bell. Some take action.

• rumor

• {Rude}

• {Membership}

• {space}

• universe

• {space}

In August 2020, The Verg e-Newspaper reported about the "Masking" app titled Fawkes. This program distorts a selfie photo and photos left in public networks a little. This tool was developed at the Sand Lab Research Institute of Chicago Research Institute.

• universe

In November 2020, the generated media released tools under the title "Anonimizer". This program creates a series of synthetic portraits based on uploaded photos. According to the website, the images are mathematically similar to your identity, look like you, but these lies. This has the ability to freeze with an attractive conclusion to lock and down a system like a clear view A. I., which selects millions of people from the public network (details about the Clear View A. I. Dispute). 。

We tested Anonimizer on November 27, 2020. However, the useless replacement balls and 40 people were far from a truly loaded portrait.

Based on January 28, 2021, Thomas Smith's interesting experience can arrange something invisible and set a simple technique.

According to his research, wearing a single opaque sunglasses mask is a powerful composition that allows you to arrange invisible things.

In this case, they lose a lot of valuable information (mouth, nose, eyes, eyebrows) to arrange a clear face comparison.

1. {space}
2. {space}

#7 Further together – towards hybridized solutions

Space} The image (form portrait) of two or more people is not connected to the reference document (such as a personal passport).

Future solutions for identification and authentication apply all the nuances of biometrics.

This creates a biometric ensemble that can provide perfect security and privacy for all members of the ecosystem.

It is based on this spirit that Tales Gem Alto runs IDCloud Fraud Prevention, which is a software for evaluating risks in the settlement industry and detecting fraud.

In this conclusion, geolocation, I. P. Address (used device), and key templates are reliable combinations for user safe authentication when using services provided by local governments.

This seventh trend is ours.

Our job is to predict it and implement it using a highly valu e-added biometric plan.

Face recognition and you.

Tales has been practicing biometrics development for about 30 years. The company has always cooperated with the best researchers, ethics, and developers of biometric applications.

This time it's your turn.

Many changes will occur in the next few months.

In fact, we cannot predict all major themes that will be seen in the near future.

Can you fill the blanks?

If you have any reports about face recognition, development, direction, questions, or if you feel that this note is needed, please leave a comment on the following.

We also welcome services to improve this note and future notebook services.