Google pays $ 5 to scan pedestrian faces, collecting necessary data for an ambitious new project

Hundreds of Google employees have been divided into small groups, spreading across major cities across the United States to implement a gift-receiving face scanning program.

Google recently opened up some interesting information about the new face unlock feature, designed specifically for the Pixel 4 - its upcoming flagship model this year. This new feature is said to be accurate and fast no less than Face ID on iPhone.

Besides some information about software from Google, there were also images leaked about hardware and code, showing that this feature has been successfully applied on Pixel 4. However, it seems functional. This has yet to function as expected by experts, and Google is stepping up several public data collection plans to improve it.

  1. Leaking image of Pixel 4: Finally, Google also 'suffered' with multi-sensor camera cluster?

Picture 1 of Google pays $ 5 to scan pedestrian faces, collecting necessary data for an ambitious new project
List of sensors, processor chips are equipped on Pixel 4

According to the report of ZDNet and Android Police, hundreds of Google employees were divided into small groups, radiating across major cities across the United States to implement a gift-receiving face scanning program. Accordingly, Google employees will 'block' on crowded streets, consult with passersby about if they agree to let Google scan their faces, they will receive a cash payment. worth 5 dollars or gift card with equal value.

According to The Verge, Google has confirmed that the company is implementing an ambitious project they call 'field research' to collect facial scanning data to improve the effectiveness of the algorithm. , thereby improving the accuracy of face recognition feature on Pixel 4 as well as many other leading hardware products.

A Google spokesperson also recently confirmed that the main purpose of the face scanning plan in the form of field data collection is to ensure that the Pixel 4's identity is operable. variety in many different situations. In fact, most biometric authentication features - including face recognition - have a poor history of biased gender and race bias. Typically, with Amazon's case, many of the company's biometric identification teams based on the artificial intelligence were forced to halt deployment due to racial bias behaviors.

  1. AI technology and face recognition are used to build shelters for stray cats

Picture 2 of Google pays $ 5 to scan pedestrian faces, collecting necessary data for an ambitious new project
Face recognition based on artificial intelligence always contains quite a lot of bias-related issues

Google certainly doesn't want what happens on Pixel 4 or any of its technology products. 'Our goal is to build features with strong performance and security. In addition, Google is also ambitious to develop a truly comprehensive biometric identification system, so field data collection is of great significance, the more data collected means synonymous with The likelihood of success of the project is warranted, 'said a Google representative.

If you want the biometric authentication feature on Pixel 4 to avoid biases as stated, Google needs to train its new algorithm on a database made up of a series of facial images. reality - as much as possible. Due to the large amount of data needed, Google, though mobilizing all of its employees to participate in the project, is not enough, so they devised a plan to reach people on the street, pay money in exchange for a nod to agree to participate in scanning face images.

  1. Face recognition features of Chinese technology companies will dominate the world in 2019

Actually this is not a new way, Apple has previously deployed the same plan and collected the necessary data. As a result, the facial recognition feature on their products is extremely stable with high accuracy. The data collection process to improve Apple's facial recognition technology was described by vice president Cynthia Hogan in 2017:

'Product accessibility for individuals of different ethnicities and ethnicities is very important to us. Face ID's algorithm will basically use neural networks developed based on training data obtained from more than one billion different face images, including IR and collected images. In studies conducted with the consent of participants. We have approached many individuals from all over the world, thus grouping into representative models according to the sex, age, ethnicity and some other important factors. In addition, we have also increased the implementation of research as needed to bring a higher level of accuracy to many different users. In addition, artificial neural networks are also trained to detect and combat fraudulent attempts, attempting to unlock a user's phone with images that have been modified or used by the mask '.

  1. Turning his back and covering his face still didn't escape the technology of identifying Chinese gait with an accurate rate of 94%.

Picture 3 of Google pays $ 5 to scan pedestrian faces, collecting necessary data for an ambitious new project
Apple's Face ID for extremely stable operation with high accuracy

As for Google, the company is pushing to collect infrared data, color and depth from each face, along with time, ambient light levels and some other relevant information. Previously, they also intended to collect location information, but after all, this factor is not necessary, so the collected location data will be deleted by Google to ensure data privacy. .

These data types allow us to understand how Google's face unlock feature works - it will actually create an in-depth diagram of your face to ensure accuracy and security. Honey - and will certainly work in a low-light environment thanks to the help of infrared cameras.

However, the way that Google handles collected data is also a lot of concern because they have had a lot of data and privacy scandals before. Google said face image data of program participants will be anonymous and kept strictly confidential, and assigned an abstract identifier. The company will keep each participant's email address separately to delete the required data. The last part is very important: anyone who participates in a Google program can request their face data to be deleted when needed.

  1. Sony is about to introduce face detection technology with a new laser sensor, far beyond Apple's Face ID

Face data will be stored for 18 months before being deleted locally. In addition, Google also said that the faces of participants will absolutely not be associated with Google ID, as well as 'encrypted and restricted access'.

After the first Pixel 4 was officially shipped, the data used to unlock the device will not be uploaded to Google servers. In the blog post announcing this feature, Google wrote that:

'Security and privacy are Pixel's core principles as well as any of our other products. Face unlock feature uses face recognition technology that is handled directly on your device, thereby ensuring that image data will never leak from your phone. Images used for face unlocking are never saved or shared with other Google services. To protect your privacy and data security, your face images will be stored securely in Pixel's built-in Titan M security chip. Similarly, Soli sensor data will also be processed locally on your phone and will never be stored or shared with other Google services'.

  1. 10 smartphones worth looking forward to the second half of 2019

Picture 4 of Google pays $ 5 to scan pedestrian faces, collecting necessary data for an ambitious new project
Pixel 4 is one of the highly anticipated premium products this year

Pixel 4 promises to be one of the flagship models that can blow a new wind into the smartphone market during the year-end shopping season, as well as the hope of the Android world before the 'expansion' of the iPhone in the high segment. level. Let's wait and see how this phone can deliver a great experience after Google's tireless efforts and user community interest.

Update 30 July 2019
Category

System

Mac OS X

Hardware

Game

Tech info

Technology

Science

Life

Application

Electric

Program

Mobile