In order to promote the development of smart cities, Shenzhen adopts face recognition boost

The development of face recognition technology is the trend of the times, and nothing can stop it from coming. The smart city project is also a project that the country is pushing forward. To this end, Shenzhen promotes the development of smart cities and incorporates face recognition technology to make smart cities more specific.

“As of the end of 2017, the Shenzhen Municipal Public Security Bureau required the construction of 20,000-way portrait recognition in the video surveillance field throughout the city.” At a meeting, Professor An Henan from Shenzhen University introduced.

He said that nowadays AI has been integrated and landed in many industries. As far as the security industry is concerned, the Shenzhen Public Security Department has increasingly used relevant systems to assist in handling cases. In the process of system construction, the construction of Shenzhen Safe City coincided with the requirements of the Ministry of Public Security. In summary, there are four points: full HD, global coverage, network-wide sharing, and full-time availability.

Professor An revealed that the Longgang Branch is the largest branch of the Shenzhen Public Security System's face recognition construction. Up to now, the second phase of the smart police has been built. The first phase of the face has been built 108 roads, the second phase has risen to 603 roads, and the related projects have built 1761 roads. Nowadays, the face recognition access to the Longgang branch business system is 2910 roads, including 250 traffic management systems that are gathered from other branches to the Longgang branch, such as joint judgments.

He emphasized at the meeting that the application of face recognition and other systems effectively solved the difficult problems that the public security had encountered in the past. In last year's Shenzhen Top Ten Outstanding Police Cases, six cases were completely dependent on the intelligent identification of portraits, and four cases were borrowed from these systems.

Later, Professor An Henan also cites a detailed case to describe it. “In the China (Shenzhen) International Cultural Industry Fair (ICIF), the Shenzhen Municipal Public Security Bureau Futian Branch achieved nearly 200,000 person-time portrait recognition through the dynamic face control system, and more than 20 former criminals. The occurrence of zero cases in the ICIF."

For the future, Professor An Henan believes that AI and other technologies will take root in the security industry. Speech recognition and face recognition will become two major directions, including security robots and other products will be well received by the market.

The following is the full text of Professor An Henan’s speech:

Shenzhen is a city where smart security applications are relatively early. Especially in the construction of a safe city, it not only surrounds the specific construction requirements of the Ministry of Public Security for the safe city, but also has some innovations and experiences in the construction of a safe city in Shenzhen.

Wisdom security has probably gone through several "nodes" from proposing to development. From seeing to seeing clearly, from seeing clearly to being able to understand. The emergence of AI has solved the problem of how the machine "understands".

Today, the construction process of Shenzhen's safe city is mainly centered on four points:

First, full HD. The future construction goal is full HD.

Second, the global coverage. There is no dead angle, especially in the dead corner, there must be a high-definition record.

Third, the entire network to share. All-network Unicom, a wider and faster information infrastructure.

Fourth, full time available. The city can effectively use the relevant equipment anytime, anywhere.

In fact, according to the above four goals, you know that the more HD front-ends, the better the results will be, but the resulting video data will be larger. As the exponential growth of these data, the human and material growth costs of servicing these data are countless. This also shows the important position and usage environment of AI in video surveillance.

We often say that such technologies are nothing more than three dimensions to interpret these problems: big data, super-computing clusters, algorithms.

In the world, including Google, Microsoft is actually some algorithms for convolutional neural networks, but from the user's point of view, we are more concerned with the calculation of big data under the cloud architecture, including the GPU cluster, give us With the export of APIs, the future security industry, whether it is enterprises or integrated units, is more concerned about the cost of use.

This feature also determines the smart application of related products in the market.

Next, I will share some data, which is some key plans of the Shenzhen Public Security Bureau for the construction of smart cities in Shenzhen.

At the end of 2017, the Shenzhen Municipal Public Security Bureau required the construction of 20,000-way portrait recognition in the video surveillance field throughout the city.

What is the construction situation so far? First of all, let's talk about the Longgang Branch. The Longgang Branch is the largest branch of the Shenzhen Public Security System's face recognition construction. As of today, the second phase of the Smart Police has been built. The first phase of the face has been built 108, and the second phase has already risen. 603 Road, related projects have been built 1761.

Nowadays, the face recognition access to the Longgang branch business system is 2910. This 2910 road also includes 250 business systems that have been collected from other branches to the Longgang branch, such as joint judgments.

So what is the effect of so many face recognition constructions in Longgang's public security business system? Shenzhen made a top ten outstanding police case evaluation at the end of last year. Among them, 6 cases are completely dependent on the intelligent identification of portraits. There are 4 cases that borrowed this system. From these data, we can see that the related equipment is built. And the system is very beneficial to the public security department to start work.

Followed by the Futian branch. The construction of Futian Branch was relatively late, and only 65 roads were built before. However, as the region is under the jurisdiction of the Shenzhen Municipal Government, including the Convention and Exhibition Center, there are many large-scale events every year, and the demand for related programs is relatively large. For example, the ICIF, this year's ICIF has installed a total of 40 related products. In this work, the system identified a total of 180,883 people, alerted 2,871 people, and arrested 20 people, realizing the occurrence of zero cases.

Having said that, in fact, we have always sang praises to AI. "Smell the body and observe its shape", the future application of artificial intelligence in the security field will mainly be reflected in two aspects:

First, speech recognition;

Second, face recognition.

The application of these technologies involves deep learning algorithms, which are mainly related to four major blocks:

First, computing resources.

I think it is more important than the algorithm. The computing resources actually guide the super-calculation. The GPU-based super-calculation is actually based on a floating-point, wide-floating operation. In the past, it may be a cluster of one thousand CPUs. Then save 200,000 images, and it takes about a month to do a convolutional iteration. For a device like this, it takes 12 years to do it 12 times.

So today, if there are only 200 GPUs, it takes only two hours to integrate 10 million images for one iteration, which makes the recognition rate revolutionized, and the recognition of the machine surpasses humans for the first time.

In this context, both scientists and entrepreneurs are eyeing this change and bringing these results into the security industry.

Second, scale effect.

The scale effect is actually the key to the problem. The crackdown of individual cases may not be enough to explain the problem, but large-scale construction may become the orientation of related products in the security industry.

Third, processing.

Data mining and analysis after system identification.

Fourth, homework.

It refers to some of the business systems associated with the security sector. In the past, a lot of data processing was done in the background. In the future, the application can be moved forward, integrated by chip algorithm, and placed at the front end of the camera. So what is the relationship between this amount of work and the amount of integration? At present, many companies can obtain a large amount of data through many channels. At present, there are many Chinese people and there are many data. But after data collection and crawling, how about the data collation?

Here, I want to interpret a story, AlphaGo may be the deepest damage to the general public, this machine can actually win.

In fact, in 1996, there was also a very sensational practice - Deep Blue, IBM introduced a supercomputer dedicated to chess, weighing 1270 kg, only 32 CPUs, the thread is parallel, 4 cores, probably More than 120 computing units.

In the same year as the Russian chess master, the result was: the machine lost, and the man won. At that time, the so-called computer and the AI ​​we talked about today were not a concept at all, and the structure was completely different. However, when the war was re-opened in 1997, Deep Blue won the chess master. At that time, some people were very scared. What should I do in the future?

Today, the deep learning-based AI architecture we have seen is a deep convolutional standard iteration. It is actually a huge computational resource behind the cost, which is just mentioned.

The latter indicators are related to our business system. One is the public security business, which not only looks at the face, but also looks at things, listens to sounds, and discriminates categories. There are many future application scenarios:

Category 1, video surveillance. This application is very large, including video access control.

Category 2, robots. Since 2015, security robots have sprung up, including today's very hot logistics sorting robots. The application of robots in the security field cannot be underestimated.

In the future, the 'opportunity' of AI in the security field will be very large and may be reflected in two aspects:

Aspect 1, video surveillance. Video surveillance is not important, and it must be understood in the future.

Aspect 2, big data and video structuring technology.

The video records unstructured data. The transformation of structured data is a very important process and a very difficult technology. Of course, there has been a lot of progress today, but it can also be changed from manual to semi-automatic; in addition, our labeling today actually has artificial factors.

As the amount of data increases, AI will enter a period of rapid development; at the same time, due to the development of AI, the dependence of various industries on AI will become larger and larger.

High Speed Blender

High Speed Blender,Ipl Laser Hair Removal,Diy Laser Hair Removal,Best Ipl Hair Removal,Epilator Ipl

SHENZHEN CHONDEKUAI TECHNOLOGY CO.LTD , https://www.siheyidz.com