Plastic surgery will cause the face recognition security check to fail?

Plastic surgery will cause the face recognition security check to fail?

---

In fact, there is no 100% accurate answer. After all, bone removal, chin pulling, nose pad, double eyelid cutting, injections... There are too many cosmetic procedures, and the failure cases are also strange.

First, micro facelift. The facial changes that it brings may not have as much changes in the expression of the person itself, such as changing a single eyelid to a double eyelid, which basically does not affect recognition.

Taking the face recognition technology of Youtu as an example, the extraction of facial features itself has relatively strong stability, and for a certain range of expression changes, age changes, and posture changes, it will not affect the recognition effect.

(For example: take out your unforgettable ID photo, it is not a problem at all to identify it)

Second, plastic surgery with relatively large movements-it depends on whether the plastic surgery area seriously affects the feature extraction.

For the AI face recognition algorithm, the importance of different face regions for identification: the facial features of the red and yellow areas are particularly important for the identification of the face.

(According to the deep learning algorithm-automatic learning to identify the most useful features)

So to put it simply, if you have to have a plastic surgery and want to not be affected by the security check, change the outer contour of the face, such as cutting the chin, the risk is relatively small, and you must be careful when you move your eyes and nose.

Reminder: Be cautious about plastic surgery. The machine has not completely replaced manual labor. In case of embarrassment that cannot pass the security check, please change the ID photo in time.
——By Tencent Youtu Lab

Based on this speculation: getting fat should not affect the recognition rate. Don't worry!

Add a few issues that everyone cares about:

1. Major changes in hair (not shaved, big beard), will it have a big impact on face recognition?
Answer: As hair, especially hair, is considered to be a very unstable feature, AI face recognition algorithm design will selectively use this feature sparingly. From the discriminative heat map visualized by the machine, it can be found that the importance of these areas (hair, chin) is not high. Compared with wearing large sunglasses and masks, the effect of hair changes, especially hair, is relatively small.

2. How will the change of age affect face recognition? What age difference has less influence?
Answer: Generally speaking, the larger the age span, the more difficult it is to identify. The children grow up fast, maybe the same every year. At this time, it may be difficult for the recognition algorithm to recognize correctly after 2-3 years. When a person reaches adulthood, the changes in facial features over time are not so obvious. In this case, the age change of 5-10 years has little effect on face recognition.

3. Can facial recognition technology use street cameras to search for suspects or find missing children like in movies? What is the difference between this and face recognition during security check?

Answer: From a technical point of view, searching for suspects or finding missing children is 1:N face recognition. I don’t know the identity of the offender or the missing child, so I need to find the identity from a huge collection.

The security face recognition is a 1:1 face verification. It only needs to determine whether the face of the on-site document holder is consistent with the face identity in the document photo.

In the movie, street cameras are used to search for suspects or missing children. With the continuous breakthroughs in face recognition technology, such scenes have almost become a reality. Although the face is not so stable and sometimes even changes a lot, the current face recognition system cannot accurately recognize a certain person immediately in a sea of ​​hundreds of millions of people, but the face retrieval system can output 10 possible Or 100 candidates have turned needle hunting on the seabed into searching for objects, using facial technology to find clues to candidates, making it possible for the original artificially impossible tasks to be completed. This is the power of technological progress.

4. In addition to the core body, what other application scenarios are there for face recognition?
Answer: We currently have a lot of Internet+ cooperation with the police. For example, searching for criminal suspects or finding missing children are all applications other than verification of face recognition.

---

Plastic surgery will cause the face recognition security check to fail?-Safeagle