Shenzhen Hongdali Machinery Co.,Ltd is professional manufacturer of Conveyor Systems, and at the same time we also supply all kinds of Conveyor Accessories, which include Conveyor Belt, conveyor roller, PE/PVC lean pipe, conveyor chain, motor, pallets, Aluminum Extrusion Profile, ball transfer unit, roller rail...The conveyor accessories can be customized. We supply the conveyor accessories with good quality and competive price, also we supply the maintain service of the conveyor systems.
Conveyor Accessories Conveyor Accessories,Conveyor Parts,Gravity Roller,Motorized Drum Shenzhen Hongdali Machinery Co.,Ltd , https://www.conveyortop.com
What is NPU? What does it subvert?
Before the big analysis of the NPU, you may have to bother listening to the author's exciting man-machine battle between AlphaGo and Li Shishi. Many people expressed surprise and awe at AlphaGo's victory, but they did not know much about its winning "weapons." In fact, the essence of AlphaGo's victory lies in its familiarity with the deep learning algorithm of the human brain's biological mechanism.
So what is deep learning? Deep learning is an artificial intelligence algorithm derived from the biomimetic study of the brain mechanism of biological humans. In the popular sense, it means that computers learn, judge, and make decisions through the deep neural network and the mechanism that simulates the human brain. Currently, such methods are used for face recognition, speech recognition, and the like. What needs to be emphasized is that deep learning requires the support of powerful computational system computing capabilities. This kind of computing capability is inseparable from the development of the processor, and is also closely related to the NPU that the author wants to talk about today.
Currently, the processor used by AlphaGo is a CPU processor that is commonly used in other fields. Although this common processor is universal, it is not professional. It is understood that in 2010, Google used only 16,000 processors to train a deep learning neural network that recognizes cat faces. Not to mention the AlphaGo that defeated humans on Go. In the future, if artificial intelligence is to use the CPU to realize a 100 billion neuron network like the human brain, it may not be a long-term solution. Therefore, a processor that can make the deep learning system miniaturized and can be used in an embedded system, the NPU, emerges.
Unlike CPUs, which are relatively laborious to run large-scale arithmetic models, the information processing capabilities of NPU processors are 100 times or even 1000 times higher than those of CPUs. According to reports, the NPU has adopted a "data-driven parallel computing" architecture that has overturned the traditional von Neumann computer architecture used by CPUs. This kind of data flow type processor greatly enhances the ratio of computing power and power consumption, and is particularly good at processing massive multimedia data of video and images, making artificial intelligence able to show its skill in embedded machine vision applications. According to Zhang Yundong, Chief Technology Officer of Zhongxing Micro, a high-tech enterprise in Zhongguancun, if the CPU's data transmission mode is likened to “a single force, a single bridge,†then the NPU will give the data a green light in the 24 lanes of the early morning. Also low power consumption.
In this sense, it means that the NPU chip is an integral part of the future of smart hardware. With it, it can be considered as a reliable and reliable computing platform.
How does NPU lead the development of video surveillance?
Of course, the transformation brought by the NPU can not only replace the CPU, but its emergence means that the development of the video surveillance industry has officially entered the era of intelligence.
As I mentioned above, the NPU is good at processing massive multimedia data such as videos and images. Therefore, the video surveillance industry will be the main area in which it can give full play to its expertise. As we all know, video surveillance is a relatively converged system, covering a wide range, coupled with the ever-increasing codec capabilities of high-definition video, leading to exponential growth and complex data in video surveillance. To "decode" these complex data, more intelligent analysis methods are needed.
It should be noted that installing the NPU chip in the camera is just like giving the "eyes" brains with IQ burst watches. It can continue to improve its IQ through learning and training, and gradually increase the identification category. After deep learning, it can record the events of interest as a digital tag in the video stream. If we are looking for a piece of information, we only need to intelligently retrieve it in the background without relying on manual video viewing. In other words, in the future, we can search for words like “long hair, glasses, and satchels†in the video as if we were searching for text.
It should be noted that, in addition to real-time search, the NPU chip can also capture the information that has already been captured in the code stream. When necessary, the NPU chip can be searched according to the feature at any time, or use the “map search†method to use a picture to similarly Find out all the maps.
From the digital age into the smart age, video surveillance contains unlimited business opportunities. With NPU chips, it's like having a skill opener "clearance card." It is understood that the current NPU has been successfully industrialized in the field of video surveillance, and the next step will be widely used in the field of embedded machine vision such as smart driving assistance, drones, and robots.
Conclusion: In the past, the industry has witnessed the development of artificial intelligence in the field of video surveillance, but unfortunately it lacks achievable hard “weapons.†With the addition of NPU's portrait, it can be said that there is a sense of help in hardware. In the future, how to design a product with more market prospects under the support of the NPU, then look at the security companies. 
What is NPU? How does it overturn video surveillance?
In recent days, the "Long March No. 7 First Flight" has smashed all major media sections. With the successful launch last night, this exciting revelry has been inspired to the highest point. Apart from being excited and proud, everyone should not overlook a major news: China's first embedded neural network processor (NPU) was born. This is a great news for the intelligent industry, and it has a transcendental significance for the field of video surveillance.