This article provides ten face recognition policy suggestions for this, roughly addressing potential face recognition risks from three levels: face data, testing technology, and authentication evaluation subjects.
Editor’s note: This article is from WeChat public account “ Tencent Research Institute ” (ID: cyberlawrc), authors Cai Xiongshan and Yuan Jun.
Top ten core suggestions
1. Limit data storage period
2. Restrict data sharing
3. Set up a face recognition logo in public places
4. Improve the accuracy of face recognition
5. Carry out independent third-party assessments
6. Reduce incidental information collection
7. Including opt-in and opt-out in business scenarios
8. Develop uniform technical standards
9. Improve certification of standardization organizations
10. Ensure data representativeness and practical testing
At present, face recognition applications continue to penetrate into daily life, and they are widely used in missing search and rescue, security upgrades, blind guidance, anti-terrorism and other fields, but at the same time, it has also caused public anxiety and doubt.
Given that there are still many uncertain risks in the application of face recognition technology, this article provides ten face recognition policy recommendations for this, roughly dealing with potential faces from the three levels of face data, testing technology, and certification evaluation subjects. Identify risks.
01 Limit data storage period
Facial Recognition refers to the static image or dynamic video of a specific scene, using existing stored facial image databases to verify and identify the identity match of a single or multiple people in a specific scene . Face recognition usually consists of three parts: face detection and segmentation in specific scenes, extraction and analysis of facial features, and matching databases to identify faces.
The long-term digital storage of face images is one of the most frightening reasons for face recognition. There is a lot of risk of misuse of this face information. There is an urgent need to change the way to cope with public anxiety. One of the ways is to set up images and SeeFrequency storage period. When in an emergency situation, some specific image data storage is really necessary.
Once the danger period has passed, there is no need to retain the facial data retained at the critical moment. Therefore, for most application scenarios, limiting the storage period of face data can balance the benefits of multiple benefits of face recognition with minimizing risks.
The specific storage time varies from scene to scene. For example, specific images compiled to cope with unexpected situations have high instantaneous value, while other situations need to be incorporated into a huge database for subsequent matching and recognition of facial features. This article introduces a machine learning model that guarantees data security-Federated Learning. This learning mode is a decentralized machine learning solution for training data, ensuring that data is only stored in the camera terminal and not transmitted to the central data center, thereby improving data security.
02 Restrict data sharing
It is worrying that the same data is used to share and flow between multiple different purposes. For example, the U.S. Vehicle Administration sells recognition images to third-party agencies for face recognition in other scenarios.
This concern is that during the sharing of face data, the data subject is completely unaware that the data is being used for other purposes, causing the informed consent mechanism to fail. Even if the data subject knows beforehand, it is likely that they will not agree to share facial recognition data to businesses. For commercial use. Therefore, if you want to share face recognition data across scenes, you must provide justification.
According to a poll by the Brookings Institution, people ’s attitudes toward face recognition vary from scene to scene. Among them, the consent rate for the use of identification technology to protect school students is as high as 41%, the agreement rate for airport security and stadiums is approximately 30%, and the lowest consent rate is used by shops to prevent theft.
03 Face recognition logo in public places
Whether it is a private entity or a public authority, when you use it to take photos, shoot videos, or collect public information for other purposes to identify faces, you should set up clear face recognition signs in public places to clearly inform people here Face recognition systems can potentially affect public compliance with public order. In the long run, not only will people’s public safety awareness of face recognition be used, but also the freedom of choice for groups who do not want to be recorded.
04 Improving face recognition accuracy
The accuracy of face recognition will first be affected by the identified objects—different ethnic groups. Face mode features can be divided into two categories: skin color features and grayscale features. Skin color, as important information of a human face, is independent of other facial details and has relative stability. However, the recognition accuracy of the white recognition system for whites is higher than that of non-whites, and the deeper the skin color, the lower the accuracy. At this time, the recognition bias will be affected.The relative stability of the skin tone is enlarged.
In addition, the incomplete and non-representative training data of ethnic groups will exacerbate the inherent bias of face recognition. When the technology is applied to public places such as law enforcement, border security, retail, and airports, there will be problems such as prejudice and discrimination of different ethnic groups. Another important factor causing the decrease in recognition accuracy is lighting.
After all, a human face as a three-dimensional object is inevitably affected by external factors such as light, shadow, and intensity. Illumination will change the relative distribution of gray levels in the face image, so the changes in the face image caused by illumination are higher than those caused by individual differences.
According to a study from Cardiff University, there have been thousands of face matching errors in Australia. It can be seen that before large-scale application in public places, it is most urgent to clarify the accuracy standard of face recognition, and the clarification of the standard can be judged by the level of influence on people’s lives. If the core rights of the people are seriously interfered in the law enforcement process, such as arrest or imprisonment, the recognition must reach the corresponding level.
05 To carry out independent third-party assessments
Introducing independent third-party assessments can boost public confidence in their face recognition products and services. Consumers want the products they buy to deliver their inherent benefits without creating other technical issues. In order to help consumers know the functions and potential hidden dangers of face products, consider establishing a star recognition system for face recognition, or following the Energy Star program jointly implemented by the US Department of Energy and the Environmental Protection Agency, to identify many faces The application is included in the scope of third-party assessment certification.
06 Reduce incidental information collection
Some face recognition applications collect massive amounts of information that are irrelevant to the main purpose, and violate the “least-enough principle”. For example, when the police carried a camcorder to the scene for investigation, they not only photographed the suspect, but also happened to photograph passersby nearby. Law enforcement agencies need not retain irrelevant information unless it is clearly shown that the evidence is relevant to the case. When the captured image is no longer of investigative value, it can be blurred or even deleted.
07 Including opt-in and opt-out in business scenarios
The so-called opt-in refers to the consent of the data subject before sharing and utilizing the face biological information involved in identification. For example, in the case where face recognition technology associates the detected name with a personal portrait and pushes a commercial to me, it is necessary to respect his right to choose and agree in advance.
In the context of global protection of personal data, data subjects have become increasingly concerned about personal privacy. As the privacy of biological information behind face recognition, it has typical identifiability, which belongs to personal biological data and should be included in the category of personal sensitive data.Being focused on protection.
In addition to the opt-in mechanism, opt-out and the right to be forgotten can also apply. In scenarios where the risk factor is low and no long-term storage of data is required, the public is given the right to choose relevant institutions to not continue to collect or data sharing is maintained at an acceptable and reasonable level. As time goes on, high-value face data may gradually become outdated, irrelevant, out-of-range, and even harmful. At this time, it is necessary to introduce the right to be forgotten, which can improve the public’s acceptance of face recognition.
08 Develop uniform technical standards
It is customary for private market players to develop technical standards to ensure product safety. Take the mobile communication technology of that year as an example. While it was still in the development stage, industry experts formulated general standards for communication, security and compatibility. All phones must meet the above technical specifications before they can be sold. Today’s face recognition technology follows the same principle. Face recognition technology should also develop international universal technical standards to ensure that face technology is secure, privacy is not violated, and people’s fears are alleviated.
As all technologies have two sides, face recognition technology also urgently needs to create “responsible face recognition.” Fortunately, the Institute of Electrical and Electronics Engineering (IEEE) and the National Institute of Standards and Technology (NIST) are developing unified technical standards to regulate related technology applications.
09 Improving Standardization Organization Certification
Enterprise system safety factors are verified by the International Organization for Standardization (ISO). For specific products of the enterprise, the ISO organization evaluates whether they meet the requirements of the regulatory rules, and third-party agencies perform compliance tests to ensure that consumers have the right to know the technical standards of the entire process.
In the United States, NIST is responsible for product technology certification. It compares the face detection results through a public database and authenticates related applications. However, there are voices criticizing NIST for relying too much on the initial data on the private website, which cannot be promoted to daily use scenarios. Wait.
Therefore, facial technology verification should combine automatic testing with manual review, improve the certification of standardized organizations, and create reliable testing and credible verification.
10 Ensure data representativeness and test practicality
In order to ensure the accuracy of face recognition, face verification, technical standards, and government compliance testing need to be based on widely representative, non-specific data. In the face of the commercialization of face recognition, the use of expressive databases for baseline testing and product certification is particularly critical.
Single-use data, such as police facial photos, cannot fully represent all groups, and the test value will be greatly reduced. In addition to the data need to haveIn addition to being representative, practical testing is particularly important to raise public concerns about facial recognition. After all, face recognition tests based on massive image information, actual application environment, and representative crowd sample groups can effectively overcome the negative impact of lighting conditions and image resolution on test accuracy.
Note: This research report comes from “10 actions that will protect people from facial recognition software” released by the Brookings Institution on October 31. This article was translated and analyzed by Cai Xiongshan and Yuan Jun of the Legal Research Center of Tencent Research Institute.
It is worrying that the same data is used to share and flow between multiple different purposes. For example, the U.S. Vehicle Administration sells recognition images to third-party agencies for face recognition in other scenarios.
This concern is that during the sharing of face data, the data subject is completely unaware that the data is being used for other purposes, causing the informed consent mechanism to fail. Even if the data subject knows beforehand, it is likely that they will not agree to share facial recognition data to businesses. For commercial use. Therefore, if you want to share face recognition data across scenes, you must provide justification.
According to a poll by the Brookings Institution, people ’s attitudes toward face recognition vary from scene to scene. Among them, the consent rate for the use of identification technology to protect school students is as high as 41%, the agreement rate for airport security and stadiums is approximately 30%, and the lowest consent rate is used by shops to prevent theft.
03 Face recognition logo in public places
Whether it is a private entity or a public authority, when you use it to take photos, shoot videos, or collect public information for other purposes to identify faces, you should set up clear face recognition signs in public places to clearly inform people here Face recognition systems can potentially affect public compliance with public order. In the long run, not only will people’s public safety awareness of face recognition be used, but also the freedom of choice for groups who do not want to be recorded.
04 Improving face recognition accuracy
The accuracy of face recognition will first be affected by the identified objects—different ethnic groups. Face mode features can be divided into two categories: skin color features and grayscale features. Skin color, as important information of a human face, is independent of other facial details and has relative stability. However, the recognition accuracy of the white recognition system for whites is higher than that of non-whites, and the deeper the skin color, the lower the accuracy. At this time, the recognition bias will be affected.The relative stability of the skin tone is enlarged.
In addition, the incomplete and non-representative training data of ethnic groups will exacerbate the inherent bias of face recognition. When the technology is applied to public places such as law enforcement, border security, retail, and airports, there will be problems such as prejudice and discrimination of different ethnic groups. Another important factor causing the decrease in recognition accuracy is lighting.
After all, a human face as a three-dimensional object is inevitably affected by external factors such as light, shadow, and intensity. Illumination will change the relative distribution of gray levels in the face image, so the changes in the face image caused by illumination are higher than those caused by individual differences.
According to a study from Cardiff University, there have been thousands of face matching errors in Australia. It can be seen that before large-scale application in public places, it is most urgent to clarify the accuracy standard of face recognition, and the clarification of the standard can be judged by the level of influence on people’s lives. If the core rights of the people are seriously interfered in the law enforcement process, such as arrest or imprisonment, the recognition must reach the corresponding level.
05 To carry out independent third-party assessments
Introducing independent third-party assessments can boost public confidence in their face recognition products and services. Consumers want the products they buy to deliver their inherent benefits without creating other technical issues. In order to help consumers know the functions and potential hidden dangers of face products, consider establishing a star recognition system for face recognition, or following the Energy Star program jointly implemented by the US Department of Energy and the Environmental Protection Agency, to identify many faces The application is included in the scope of third-party assessment certification.
06 Reduce incidental information collection
Some face recognition applications collect massive amounts of information that are irrelevant to the main purpose, and violate the “least-enough principle”. For example, when the police carried a camcorder to the scene for investigation, they not only photographed the suspect, but also happened to photograph passersby nearby. Law enforcement agencies need not retain irrelevant information unless it is clearly shown that the evidence is relevant to the case. When the captured image is no longer of investigative value, it can be blurred or even deleted.
07 Including opt-in and opt-out in business scenarios
The so-called opt-in refers to the consent of the data subject before sharing and utilizing the face biological information involved in identification. For example, in the case where face recognition technology associates the detected name with a personal portrait and pushes a commercial to me, it is necessary to respect his right to choose and agree in advance.
In the context of global protection of personal data, data subjects have become increasingly concerned about personal privacy. As the privacy of biological information behind face recognition, it has typical identifiability, which belongs to personal biological data and should be included in the category of personal sensitive data.Being focused on protection.
In addition to the opt-in mechanism, opt-out and the right to be forgotten can also apply. In scenarios where the risk factor is low and no long-term storage of data is required, the public is given the right to choose relevant institutions to not continue to collect or data sharing is maintained at an acceptable and reasonable level. As time goes on, high-value face data may gradually become outdated, irrelevant, out-of-range, and even harmful. At this time, it is necessary to introduce the right to be forgotten, which can improve the public’s acceptance of face recognition.
08 Develop uniform technical standards
It is customary for private market players to develop technical standards to ensure product safety. Take the mobile communication technology of that year as an example. While it was still in the development stage, industry experts formulated general standards for communication, security and compatibility. All phones must meet the above technical specifications before they can be sold. Today’s face recognition technology follows the same principle. Face recognition technology should also develop international universal technical standards to ensure that face technology is secure, privacy is not violated, and people’s fears are alleviated.
As all technologies have two sides, face recognition technology also urgently needs to create “responsible face recognition.” Fortunately, the Institute of Electrical and Electronics Engineering (IEEE) and the National Institute of Standards and Technology (NIST) are developing unified technical standards to regulate related technology applications.
09 Improving Standardization Organization Certification
Enterprise system safety factors are verified by the International Organization for Standardization (ISO). For specific products of the enterprise, the ISO organization evaluates whether they meet the requirements of the regulatory rules, and third-party agencies perform compliance tests to ensure that consumers have the right to know the technical standards of the entire process.
In the United States, NIST is responsible for product technology certification. It compares the face detection results through a public database and authenticates related applications. However, there are voices criticizing NIST for relying too much on the initial data on the private website, which cannot be promoted to daily use scenarios. Wait.
Therefore, facial technology verification should combine automatic testing with manual review, improve the certification of standardized organizations, and create reliable testing and credible verification.
10 Ensure data representativeness and test practicality
In order to ensure the accuracy of face recognition, face verification, technical standards, and government compliance testing need to be based on widely representative, non-specific data. In the face of the commercialization of face recognition, the use of expressive databases for baseline testing and product certification is particularly critical.
Single-use data, such as police facial photos, cannot fully represent all groups, and the test value will be greatly reduced. In addition to the data need to haveIn addition to being representative, practical testing is particularly important to raise public concerns about facial recognition. After all, face recognition tests based on massive image information, actual application environment, and representative crowd sample groups can effectively overcome the negative impact of lighting conditions and image resolution on test accuracy.
Note: This research report comes from “10 actions that will protect people from facial recognition software” released by the Brookings Institution on October 31. This article was translated and analyzed by Cai Xiongshan and Yuan Jun of the Legal Research Center of Tencent Research Institute.
The accuracy of face recognition will first be affected by the identified objects—different ethnic groups. Face mode features can be divided into two categories: skin color features and grayscale features. Skin color, as important information of a human face, is independent of other facial details and has relative stability. However, the recognition accuracy of the white recognition system for whites is higher than that of non-whites, and the deeper the skin color, the lower the accuracy. At this time, the recognition bias will be affected.The relative stability of the skin tone is enlarged.
In addition, the incomplete and non-representative training data of ethnic groups will exacerbate the inherent bias of face recognition. When the technology is applied to public places such as law enforcement, border security, retail, and airports, there will be problems such as prejudice and discrimination of different ethnic groups. Another important factor causing the decrease in recognition accuracy is lighting.
After all, a human face as a three-dimensional object is inevitably affected by external factors such as light, shadow, and intensity. Illumination will change the relative distribution of gray levels in the face image, so the changes in the face image caused by illumination are higher than those caused by individual differences.
According to a study from Cardiff University, there have been thousands of face matching errors in Australia. It can be seen that before large-scale application in public places, it is most urgent to clarify the accuracy standard of face recognition, and the clarification of the standard can be judged by the level of influence on people’s lives. If the core rights of the people are seriously interfered in the law enforcement process, such as arrest or imprisonment, the recognition must reach the corresponding level.
05 To carry out independent third-party assessments
Introducing independent third-party assessments can boost public confidence in their face recognition products and services. Consumers want the products they buy to deliver their inherent benefits without creating other technical issues. In order to help consumers know the functions and potential hidden dangers of face products, consider establishing a star recognition system for face recognition, or following the Energy Star program jointly implemented by the US Department of Energy and the Environmental Protection Agency, to identify many faces The application is included in the scope of third-party assessment certification.
06 Reduce incidental information collection
Some face recognition applications collect massive amounts of information that are irrelevant to the main purpose, and violate the “least-enough principle”. For example, when the police carried a camcorder to the scene for investigation, they not only photographed the suspect, but also happened to photograph passersby nearby. Law enforcement agencies need not retain irrelevant information unless it is clearly shown that the evidence is relevant to the case. When the captured image is no longer of investigative value, it can be blurred or even deleted.
07 Including opt-in and opt-out in business scenarios
The so-called opt-in refers to the consent of the data subject before sharing and utilizing the face biological information involved in identification. For example, in the case where face recognition technology associates the detected name with a personal portrait and pushes a commercial to me, it is necessary to respect his right to choose and agree in advance.
In the context of global protection of personal data, data subjects have become increasingly concerned about personal privacy. As the privacy of biological information behind face recognition, it has typical identifiability, which belongs to personal biological data and should be included in the category of personal sensitive data.Being focused on protection.
In addition to the opt-in mechanism, opt-out and the right to be forgotten can also apply. In scenarios where the risk factor is low and no long-term storage of data is required, the public is given the right to choose relevant institutions to not continue to collect or data sharing is maintained at an acceptable and reasonable level. As time goes on, high-value face data may gradually become outdated, irrelevant, out-of-range, and even harmful. At this time, it is necessary to introduce the right to be forgotten, which can improve the public’s acceptance of face recognition.
08 Develop uniform technical standards
It is customary for private market players to develop technical standards to ensure product safety. Take the mobile communication technology of that year as an example. While it was still in the development stage, industry experts formulated general standards for communication, security and compatibility. All phones must meet the above technical specifications before they can be sold. Today’s face recognition technology follows the same principle. Face recognition technology should also develop international universal technical standards to ensure that face technology is secure, privacy is not violated, and people’s fears are alleviated.
As all technologies have two sides, face recognition technology also urgently needs to create “responsible face recognition.” Fortunately, the Institute of Electrical and Electronics Engineering (IEEE) and the National Institute of Standards and Technology (NIST) are developing unified technical standards to regulate related technology applications.
09 Improving Standardization Organization Certification
Enterprise system safety factors are verified by the International Organization for Standardization (ISO). For specific products of the enterprise, the ISO organization evaluates whether they meet the requirements of the regulatory rules, and third-party agencies perform compliance tests to ensure that consumers have the right to know the technical standards of the entire process.
In the United States, NIST is responsible for product technology certification. It compares the face detection results through a public database and authenticates related applications. However, there are voices criticizing NIST for relying too much on the initial data on the private website, which cannot be promoted to daily use scenarios. Wait.
Therefore, facial technology verification should combine automatic testing with manual review, improve the certification of standardized organizations, and create reliable testing and credible verification.
10 Ensure data representativeness and test practicality
In order to ensure the accuracy of face recognition, face verification, technical standards, and government compliance testing need to be based on widely representative, non-specific data. In the face of the commercialization of face recognition, the use of expressive databases for baseline testing and product certification is particularly critical.
Single-use data, such as police facial photos, cannot fully represent all groups, and the test value will be greatly reduced. In addition to the data need to haveIn addition to being representative, practical testing is particularly important to raise public concerns about facial recognition. After all, face recognition tests based on massive image information, actual application environment, and representative crowd sample groups can effectively overcome the negative impact of lighting conditions and image resolution on test accuracy.
Note: This research report comes from “10 actions that will protect people from facial recognition software” released by the Brookings Institution on October 31. This article was translated and analyzed by Cai Xiongshan and Yuan Jun of the Legal Research Center of Tencent Research Institute.
Some face recognition applications collect massive amounts of information that are irrelevant to the main purpose, and violate the “least-enough principle”. For example, when the police carried a camcorder to the scene for investigation, they not only photographed the suspect, but also happened to photograph passersby nearby. Law enforcement agencies need not retain irrelevant information unless it is clearly shown that the evidence is relevant to the case. When the captured image is no longer of investigative value, it can be blurred or even deleted.
07 Including opt-in and opt-out in business scenarios
The so-called opt-in refers to the consent of the data subject before sharing and utilizing the face biological information involved in identification. For example, in the case where face recognition technology associates the detected name with a personal portrait and pushes a commercial to me, it is necessary to respect his right to choose and agree in advance.
In the context of global protection of personal data, data subjects have become increasingly concerned about personal privacy. As the privacy of biological information behind face recognition, it has typical identifiability, which belongs to personal biological data and should be included in the category of personal sensitive data.Being focused on protection.
In addition to the opt-in mechanism, opt-out and the right to be forgotten can also apply. In scenarios where the risk factor is low and no long-term storage of data is required, the public is given the right to choose relevant institutions to not continue to collect or data sharing is maintained at an acceptable and reasonable level. As time goes on, high-value face data may gradually become outdated, irrelevant, out-of-range, and even harmful. At this time, it is necessary to introduce the right to be forgotten, which can improve the public’s acceptance of face recognition.
08 Develop uniform technical standards
It is customary for private market players to develop technical standards to ensure product safety. Take the mobile communication technology of that year as an example. While it was still in the development stage, industry experts formulated general standards for communication, security and compatibility. All phones must meet the above technical specifications before they can be sold. Today’s face recognition technology follows the same principle. Face recognition technology should also develop international universal technical standards to ensure that face technology is secure, privacy is not violated, and people’s fears are alleviated.
As all technologies have two sides, face recognition technology also urgently needs to create “responsible face recognition.” Fortunately, the Institute of Electrical and Electronics Engineering (IEEE) and the National Institute of Standards and Technology (NIST) are developing unified technical standards to regulate related technology applications.
09 Improving Standardization Organization Certification
Enterprise system safety factors are verified by the International Organization for Standardization (ISO). For specific products of the enterprise, the ISO organization evaluates whether they meet the requirements of the regulatory rules, and third-party agencies perform compliance tests to ensure that consumers have the right to know the technical standards of the entire process.
In the United States, NIST is responsible for product technology certification. It compares the face detection results through a public database and authenticates related applications. However, there are voices criticizing NIST for relying too much on the initial data on the private website, which cannot be promoted to daily use scenarios. Wait.
Therefore, facial technology verification should combine automatic testing with manual review, improve the certification of standardized organizations, and create reliable testing and credible verification.
10 Ensure data representativeness and test practicality
In order to ensure the accuracy of face recognition, face verification, technical standards, and government compliance testing need to be based on widely representative, non-specific data. In the face of the commercialization of face recognition, the use of expressive databases for baseline testing and product certification is particularly critical.
Single-use data, such as police facial photos, cannot fully represent all groups, and the test value will be greatly reduced. In addition to the data need to haveIn addition to being representative, practical testing is particularly important to raise public concerns about facial recognition. After all, face recognition tests based on massive image information, actual application environment, and representative crowd sample groups can effectively overcome the negative impact of lighting conditions and image resolution on test accuracy.
Note: This research report comes from “10 actions that will protect people from facial recognition software” released by the Brookings Institution on October 31. This article was translated and analyzed by Cai Xiongshan and Yuan Jun of the Legal Research Center of Tencent Research Institute.
It is customary for private market players to develop technical standards to ensure product safety. Take the mobile communication technology of that year as an example. While it was still in the development stage, industry experts formulated general standards for communication, security and compatibility. All phones must meet the above technical specifications before they can be sold. Today’s face recognition technology follows the same principle. Face recognition technology should also develop international universal technical standards to ensure that face technology is secure, privacy is not violated, and people’s fears are alleviated.
As all technologies have two sides, face recognition technology also urgently needs to create “responsible face recognition.” Fortunately, the Institute of Electrical and Electronics Engineering (IEEE) and the National Institute of Standards and Technology (NIST) are developing unified technical standards to regulate related technology applications.
09 Improving Standardization Organization Certification
Enterprise system safety factors are verified by the International Organization for Standardization (ISO). For specific products of the enterprise, the ISO organization evaluates whether they meet the requirements of the regulatory rules, and third-party agencies perform compliance tests to ensure that consumers have the right to know the technical standards of the entire process.
In the United States, NIST is responsible for product technology certification. It compares the face detection results through a public database and authenticates related applications. However, there are voices criticizing NIST for relying too much on the initial data on the private website, which cannot be promoted to daily use scenarios. Wait.
Therefore, facial technology verification should combine automatic testing with manual review, improve the certification of standardized organizations, and create reliable testing and credible verification.
10 Ensure data representativeness and test practicality
In order to ensure the accuracy of face recognition, face verification, technical standards, and government compliance testing need to be based on widely representative, non-specific data. In the face of the commercialization of face recognition, the use of expressive databases for baseline testing and product certification is particularly critical.
Single-use data, such as police facial photos, cannot fully represent all groups, and the test value will be greatly reduced. In addition to the data need to haveIn addition to being representative, practical testing is particularly important to raise public concerns about facial recognition. After all, face recognition tests based on massive image information, actual application environment, and representative crowd sample groups can effectively overcome the negative impact of lighting conditions and image resolution on test accuracy.
Note: This research report comes from “10 actions that will protect people from facial recognition software” released by the Brookings Institution on October 31. This article was translated and analyzed by Cai Xiongshan and Yuan Jun of the Legal Research Center of Tencent Research Institute.
In order to ensure the accuracy of face recognition, face verification, technical standards, and government compliance testing need to be based on widely representative, non-specific data. In the face of the commercialization of face recognition, the use of expressive databases for baseline testing and product certification is particularly critical.
Single-use data, such as police facial photos, cannot fully represent all groups, and the test value will be greatly reduced. In addition to the data need to haveIn addition to being representative, practical testing is particularly important to raise public concerns about facial recognition. After all, face recognition tests based on massive image information, actual application environment, and representative crowd sample groups can effectively overcome the negative impact of lighting conditions and image resolution on test accuracy.
Note: This research report comes from “10 actions that will protect people from facial recognition software” released by the Brookings Institution on October 31. This article was translated and analyzed by Cai Xiongshan and Yuan Jun of the Legal Research Center of Tencent Research Institute.