Calendar
resents face images as a distribution rather than a point estimation. With Probabilistic Face Embedding, the mean of the distribution represents the best estimation of the facial fea- tures in the image. The variance in the distribution represents the uncertain- ty.
Is face recognition reliable?
Face recognition accuracy varies de- pending on the algorithm and the conditions. The data shows that many of the state-of-the-art methods per- form relatively well under constrained settings. On the other hand, there is a substantial drop in accuracy when tested in unconstrained settings. Acceptable accuracy levels are all a matter of scale and context. AI and Facial Recognition technologies, are used in the banking, financial services and insurance industries. CyberLink has introduced software solutions Fa- ceMe Health for pandemic control, and FaceMe Security for access con- trol, authentication and security. Those companies deploying Facial Recognition in the context of the pan- demic will probably want to know what the technology can be used for once the pandemic wanes. FaceMe eKYC and fintech solutions offer the finance and insurance industries the ability to conduct secure, real-time biometric verification and identity recognition. The technology offers financial institu- tions the ability to match a person’s photo to their ID in person or virtually, protecting customers against identity fraud whether they’re at a branch, or accessing their accounts from home. The FaceMe eKYC and fintech solution can be applied for identity verification, or second factor authentication, in opening bank accounts, applying for credit, conducting ATM transactions or mobile banking, buying insurance services, and using remote customer service. FaceMe facial recognition is described as one of the most accurate methods of identity verification with up to 99.7% accuracy rate.
Case Study of Facial-Recognition payment in China
Recently, facial-recognition payment (FRP, or Scan the face to pay. Has gained popularity in China as a new digital-payment method at physical stores. Facial-recognition payment is convenient, but users were concerned and confused after using it for the first time. Better onboarding experiences can help relieve concerns and form factual mental models.
China already has widespread mobile payment as a primary method of mak- ing purchases. In some places, cash isn’t accepted — only mobile pay- ment. A Chinese shopper can leave home without her purse, as long as she has her phone.
This wallet-free reality is largely due
to QR-code scanning: people scan the QR code of a shop and pay the amount of the order on their phones. Facial recognition takes this process a step further — you don’t even need your phone, just your face. Let’s look at the example of a bak- ery called WeiDuoMei. A specialised touchscreen device is placed at the checkout counter of the bakery. It’s similar in size to an iPad Mini. Pro- duced by either Ali or Tencent (which are the two companies that monop- olise the mobile-payment market in China), the device has access to the facial-information database of the par- ent company. So, if Tencent has your picture, you will be in its facial-infor- mation database, and the machine will be able to find you.
The cashier enters the amount of money for the order, which is then shown on the device screen. A user just needs 10 seconds and two taps to finish the payment: The user taps Pay with Face Recognition on the screen of the device. The device scans the face and recognizes the user. The user then taps Confirm Payment and is done. A user tapped Pay with Face Recogni- tion and then Confirm Payment to pay At first glance, facial recognition pay- ment might seem very similar to the process of using Apple Pay on an iP- hone: the newer iPhones (iPhone X and newer) use “Face ID” to check identi- ty and approve payment. However, the key difference here is the reliance on a device. To use Apple Pay via Face ID, users must own a late-model iPhone, must have it with them, and have it at least partially charged. To use FRP in China, users only need to have a vali- dated WeChat or Ali Pay account and an associated bank account, no matter
18
what types of phones they have (or if they have a phone with them at all). This technology enables a much larg- er audience to pay with their faces without worrying about the compati- bility of their devices. While all new iPhone and Pixel models have face recognition, these devices are not that common in China. In fact, the most popular and middle-level smartphones in China don’t have facial-recognition authentication and still use finger- print or passwords for authentication. Thus, FRP allows even users who own old-fashioned phones to pay with their faces. This new method has been adopt- ed in many physical stores in China: clothing stores, grocery stores, vend- ing machines, and checkpoints of sub- way stations. It seems cool and conve- nient, because it can reduce checkout time and the hassle of pulling out mo- bile phones.
All participants, but especially new users, praised the convenience of FRP. Even for first-time users, the method reduced the total payment time to around 15–20 seconds. One 35-year- old male, new to this process, couldn’t believe it happened so fast. When he finished the payment, he asked the ca- shier twice “So is it done?” After being repeatedly reassured that he had paid for his croissant, he left.
Probabilistic Face Embedding is a rela- tively new technique which represents face images as a distribution rather than a point estimation.
Feature
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30
