info@elefinetech.com
Tel:+8613534079085
The small “health code” on the screen of the mobile phone played a huge role in the prevention and control of the epidemic.
As of December 18, 2020, the cumulative number of health codes has exceeded 20 billion people, covering 1 billion people, and the cumulative number of visits has exceeded 50 billion. Recently, The Paper reporter interviewed Professor Ma Lizhuang, the leader of the facial recognition technology team behind the health code at the World Artificial Intelligence Conference.
Professor Ma Lizhuang
Ma Lizhuang is a distinguished professor of Shanghai Jiaotong University, deputy dean of the Institute of Artificial Intelligence, a distinguished professor of East China Normal University (concurrently); director of the Chinese Society of Graphics and Graphics, director of the Animation and Digital Entertainment Committee, director of the Chinese Society of Artificial Intelligence, intelligent creativity And Deputy Director of the Digital Art Committee.
The Ma Lizhuang team focuses on computer graphics and computer vision technology. The research and application results of the team’s “Character Visual Understanding Technology and Application for Complex Scenes” won the 2020 Shanghai Science and Technology Progress Award Special Prize in May this year.
The project team has built major applications in health codes, resumption of work and production, AI tracing, Internet finance, etc., which have generated over 10 billion economic benefits and promoted the progress of artificial intelligence-related industries; and obtained 165 authorized invention patents and software 15 copyrights, 1 IEEE international standard, 8 national standards, published 290 high-level papers (70 papers included in CCF-A), and achieved outstanding innovation results.
“Face recognition technology involves a kind of biological characteristics, and more standards are needed to conduct related research while meeting personal privacy.” Professor Ma Lizhuang said, “We can also use computer vision for medical imaging. For example. For the pathological slices we are doing, we cooperate with the Children’s Medical Research Institute to identify cancer cells at different scales. There is also a’face recognition’ to see a doctor.”
“Artificial intelligence must be people-oriented and serve people.” Ma Lizhuang emphasized that under the premise of standardized and reasonable use, face recognition, like other artificial intelligence technologies, will definitely make our lives better and more convenient.
Regarding the future development of this technology, Ma Lizhuang said, “I think the two directions that are more optimistic are service robots. In the future, they may be in the form of human beings. Behavior, capable of self-control and decision-making, and even emotion, will enter millions of households in the future; the second is augmented reality and virtual reality in entertainment.”
Ma Lizhuang: One is the health code applied during the epidemic. When people log in to verify the live body detection and face comparison, the health code can be obtained after passing, so as to ensure the authenticity of the user’s identity. The remote identity verification based on the living body of the light is the core technology of the health code. As of December 18, 2020, the cumulative number of health codes has exceeded 20 billion people, covering 1 billion people, and the cumulative number of visits has exceeded 50 billion.
This technology also helps resumption of work and production. Through its application to the “national government service platform”, it has helped a large number of business owners and individual industrial and commercial households handle livelihood matters such as postponement of social insurance and rent reduction and exemption online, thereby contributing to the prevention and control of the epidemic and embodies it. The team’s scientific and technological pioneering role in the fight against the epidemic.
In the financial sector, we have built the world’s first facial remote identity verification system for commercial use in banking systems, and applied facial verification technology to remote bank account opening for the first time in China. At present, the remote face core system has been widely used in many application scenarios such as WeChat payment and WeBank.
Our cross-age face recognition technology can help retrieve missing persons and trafficked persons. Because the facial shape of infants and young children changes significantly with age, especially after more than ten years, the face changes drastically. The cross-temporal face retrieval and analysis system constructed by our project team retrieved 1,706 missing or trafficked persons from tens of millions of people. In response to the abduction of infants and toddlers in the tracing scene, with only a few fuzzy photos of infants and toddlers, the project team used the pioneering cross-age face recognition technology to help rescue 15 children who were abducted for more than 10 years.
Ma Lizhuang: There are several aspects, one is more accurate face recognition technology, including the combination of expressions and behaviors. In fact, the human face is divided into many muscles, called action units. For example, several muscles near the mouth are moving when talking, and the upper muscles when smiling. This is a sub-regional action unit. Combined with facial expressions, it can be used for education, customs character identification, and even micro-expressions recognition can be applied to detectives and spy departments.
The other is that there are various attack techniques, such as squeezing the face and then synthesizing it. The attacks that face recognition needs to prevent include photos, masks, especially silicone masks, and digital model attacks. I estimate that deep face security technology may develop very quickly in the future.
Of course, face recognition technology involves a kind of biological characteristics, and more specifications are required, and relevant research should be carried out while satisfying personal privacy.
I think the broader face recognition technology is the visual understanding of characters, including behaviors and scenes.
We can use computer vision for medical imaging. For example, we are doing image analysis of pathological slices. We cooperate with the Institute of Children’s Medical Research to identify cancer cells at different scales; there is also “face recognition to see a doctor.”
“Refreshing face to see a doctor” is completely scientific. The growth of the human face (craniofacial) is determined by genes. The craniofacial region is determined by about 7,000 genes, of which more than 6,000 genes are related to the appearance of the face. If a certain gene is missing, it will be reflected on the human face. We are already using this technology to screen out thousands of diseases such as Down syndrome and Progeria.
“Recognizing the face and seeing a doctor” makes it take two days or even two months to determine the disease. The result is detected in a few seconds, and the doctor can make a little judgment and basically be able to give the diagnosis result in real time. This helps many patients because they can only be treated after they are diagnosed. The biggest role here is screening. Originally, there were more than 6000 kinds of inspections, but now usually only one or two inspections are required, which greatly reduces time and cost and serves patients.
In the future, we will do more multi-modal face recognition technology, these applications are very extensive, such as entertainment, autonomous driving, etc. For example, micro-expression is a small behavioral action, which is related to the application of VR and AR. Now young people like virtual characters or avatars, that is, you have a character in the virtual world, this character looks just like you, and your society and friends in the virtual world.
I’m talking about making life better. Artificial intelligence must be people-oriented and serve people.
Ma Lizhuang: That’s for sure. I think the technology itself is not harmful, and there will be problems if it is used unreasonably. Therefore, standards, policies and regulations must be established to allow the technology to be used reasonably.
Standard use and reasonable use must be the prerequisite. Under this premise, I think face recognition and artificial intelligence will definitely make our lives better and more convenient.
Ma Lizhuang: I think China’s artificial intelligence should have a fairly high status in the international arena. It should be second only to the United States. I don’t know whether it is reasonable to say that.
And our face recognition technology is the world’s first, I am very confident about this, because our policy environment has created a relatively free research atmosphere for us. Especially in the deep security technology of human face, we have opened up a small field here, and we are leading in this field. Foreign countries have made this data more difficult to obtain due to various restrictions.
While protecting privacy and preventing data from being misused, of course, the more convenient it is to use data, the better it can serve people. For example, if you don’t tell others your location when you take a taxi, you say it’s privacy, and you ask the car to pick you up at the same time. Isn’t that a joke?
Ma Lizhuang: In the short term, I think there is no shortage of talents, because the explosive growth of artificial intelligence has passed and is now in a state of steady growth.
From the perspective of talents, since artificial intelligence is a large branch of computers abroad, of course it penetrates into various application fields. With the upsurge of artificial intelligence, there are indeed many people from other directions and different professions pouring into this industry, causing a situation of uneven talents. In other words, the number of talents has increased dramatically, but the quality may need to be improved.
In colleges and universities, the education of artificial intelligence still needs to be improved. At present, there is no independent subject for artificial intelligence undergraduates. However, there is still considerable controversy as to whether it should be set up as an independent subject and later become an artificial intelligence department like the computer department.
Artificial intelligence is indeed different from other disciplines. The biggest difference is that artificial intelligence is applied to different industries, unlike some technologies where the application fields are relatively concentrated. Artificial intelligence, like the “magic oil”, can empower various industries.
For example, in the smart city construction I just mentioned, the direct role of artificial intelligence is the most obvious. Others such as industrial testing. For example, product inspection on the assembly line was originally manual. Now we are trying to use computer vision to automatically identify product defects, which can greatly reduce labor costs. This market demand is also very urgent.
From a development perspective, I think there are two more optimistic directions. One is the service robot. In the future, it may be in the form of a human being. It can recognize you as the owner, can communicate with you, understand your behavior, and be able to control and make decisions. Even with emotion, it will enter thousands of households in the future; the second is augmented reality and virtual reality in entertainment.
Ma Lizhuang: Animation is similar to virtual reality. In fact, animation requires a higher level of intelligence. With the ups and downs of the plot, the personality of the actor will also develop up and down. This development process actually fully reflects the human intelligence. To simulate this development, intelligent roles, intelligent behavior and personality development are required.
For example, if a character’s personality is positive, cheerful, and positive, then he will have a basic logic when dealing with various events.
The intelligence of artificial intelligence roles is reflected in many aspects. For example, virtual characters can think intelligently. For example, it can read the script itself. Enter a story such as “In the sunny weather, the frog jumps into the river”. Artificial intelligence needs to understand: the protagonist is a frog, its behavior is to jump into a river, and its environment is sunny weather and a river. This requires natural language understanding. Then there is the personality development of the virtual character. It needs to be intelligent and able to perform its own interpretation according to the script, rather than artificially controlling its movement.
Therefore, it is very interesting to set the development of the characters according to different plots and theme ideas.
Download Brochure