▲ NVIDIA Global Vice President, Tegra Division General Manager Deepu Talla
Big Brother is watching you!
The streets, hundreds of millions of cameras are watching your every move. Data show that by 2020, the number of cameras will reach 1 billion. These cameras work 24 hours a day, seven days a week, and continue to generate massive amounts of data.
In the past, when you become a prey to the police, they have to drag the night to monitor the video progress bar to find you swept by the shadow; now, they will effortlessly among the people to accurately identify you, regardless of your gender , Dress or age & hellip; & hellip;
NVIDIA, which specializes in games, graphics and other areas, is armed with these cameras, while using depth learning and artificial intelligence to deal with large data to reduce the workload of the police.
(AI Computing Company) NVIDIA has been deployed in the field of security layout, to be exact, mainly concentrated in the "Intelligent Video Analysis (IVA)" field.... This is part of the NVIDIA artificial intelligence urban vision, which also includes intelligent driving, intelligent health care, and intelligent industrial and other fields.
▲ NVIDIA has become an "artificial intelligence computing company" (AI Computing Company)
"Artificial intelligence is like a new form of power in the future, and every industry has to do with it," Deepu Talla, NVIDIA's global vice president and general manager, Tegra's business unit, believes in.
Before Artificial Intelligence and in-depth learning were applied to intelligent video analytics, people were doing such things manually, both in terms of accuracy and speed.
In the past, traditional computer programming with a dual thinking work, the highest accuracy can only be done 60% to 80%, which is the world's leading companies, it took 20 years to do. Now, with Artificial Intelligence and in-depth learning, it takes just over a year to achieve higher accuracy than ever before.
▲ Comparison of the accuracy rate (gray columnar) of the traditional recognition technology and the recognition accuracy of the artificial intelligence (green columnar)
Accordingly, due to the rapid development of GPU technology, the rate of recognition and analysis will be greatly improved. This will depend on NVIDIA's powerful depth-of-learning and data-processing computing platform. Earlier, they launched the "data center into the chassis" - mdash; - supercomputer DGX - 1, fully capable.
So NVIDIA in the end how to do?
NVIDIA has introduced and introduced an end-to-end (Training to Inference, back-end server to front-end camera) depth learning platform for public security, video surveillance and other fields.
▲ NVIDIA end-to-end depth of learning computing platform
The first step in artificial intelligence is to first create a neural network, and then use this neural network to train large amounts of data, the main use is DGX-1; then the training data into the cloud and data center, which is mainly Using Tesla P4 & amp; P40; Of course, these data can also be deployed to the VCR and camera, where the main use is the Jetson TX1 computing platform.
Jetson TX1 is NVIDIA's an embedded visual computing system, in March this year, officially listed in the Chinese market, the system is based on NVIDIA's Tegra processor to build. In training, recognition and analysis of performance is very good.
Proceed as follows:
1, the creation of neural networks;
2, data training;
3, the data uploaded to the data center and the cloud, you can also deploy the data to the VCR and camera;
4, identification: simply speaking, is to monitor the camera to capture the video information input system, the system will automatically separate all people (including face, clothing and even back) and all cars (including license plate, color, year).
At present, NVIDIA security hardware in the domestic manufacturers, many of them are using this Jetson TX1 hardware to build. For example, as the Structured technology can only be based on this platform to build the server.
NVIDIA another partner Hikvision as the beginning of the year launched a "blade server", which is based on 16 Jetson TX1 processor, compared to the configuration of the 21 CPU equivalent performance server, the blade needs to be occupied Space is only 1/20, the power consumption is only 1/10.
It is relying on end-to-end depth learning platform, in the field of public safety, product solutions become more intelligent, video surveillance to do better.
As to why NVIDIA to enter the field of security IVA, Deepu Talla to Lei Feng Network (public number: Lei Feng Network) said, on the one hand to maintain personal safety feelings; Two areas with the depth of learning and then appropriate. Moreover, the current analysis of intelligent video analysis to do after the forecast and analysis for the blind spot, so the need for in-depth learning and artificial intelligence to do such things.
At the same time, NVIDIA is an artificial intelligence computing company, its artificial intelligence computing platform no matter in which applications, are the most preferred platform. Second, from the solution level, NVIDIA provides solutions from training to reasoning, but also can provide from the cloud to the data center to the server to the end of such a solution. Moreover, the future NVIDIA will continue to the next generation of GPU technology investment.
In Deepu Talla's vision, the IVA market, NVIDIA hopes to conclude more partnerships in the IVA market, and hope to make existing partners more in-depth cooperation, for example, with Hikvision and NVIDIA cooperation It is very deep, such as training servers and cameras in this regard.
At present, NVIDIA has nine domestic security and security hardware vendors to cooperate, At the Expo, their servers and camera products are on display.
▲ NVIDIA and nine domestic security hardware companies to cooperate
Artificial intelligence and depth of learning are profoundly changing the video surveillance industry, so that we are in this world, have a better level of public safety. We have entered into a stage of the use of artificial intelligence more and more deep.
NVIDIA's GPU is the depth of the learning process is accelerating at the same time, in the domestic security industry, NVIDIA is also actively developing partners, hoping to dominate in this industry.
Undoubtedly, Big Brother is also more and more with artificial intelligence.