The new tens of billions of dollars market is coming soon, there is a quick way to enter
According to Omdia's 2020 report, the total size of the AI ​​inferred market in 2020 is 5 billion U.S. dollars and is expected to reach 7 billion U.S. dollars this year. Xilinx predicts that the AI ​​inferred market will grow at a compound annual growth rate of 35% in the next few years. It will exceed 10 billion U.S. dollars in two years, and even more than 20 billion U.S. dollars in 5 years.
 

Figure: AI infers that the market will grow at a compound annual growth rate of 35% (Source: Xilinx)
AI infers major opportunities in the market
So, where are the main opportunities for this new tens of billions of dollars market? According to public market analysis reports and Xilinx’s predictions, the market has two main types of applications. One is image and video-related applications, accounting for about 55% of the market; the other is voice-related applications, including natural text. Processing and recommendation applications account for about 45% of the market. In other words, video analytics will be the fastest-growing segment in the AI ​​inference market.

For video analysis, customer needs are mainly focused on real-time analysis of video streams, extracting useful features, and using this information in various practical application scenarios. For example, in the following four typical application scenarios, video analysis can implement a variety of different functions.

In smart retail application scenarios, smart cashier systems, smart shelf systems, and people flow analysis systems in stores all require video analysis; in smart parking application scenarios, the parking lot can be intelligently managed through video analysis , Including parking space analysis, vehicle parking status analysis, license plate recognition, and vehicle positioning; in smart city application scenarios, the flow of pedestrians and vehicles needs to be analyzed to solve traffic congestion. There are also intelligent inspections of key infrastructure such as electricity and gas pipelines. Due to the wide range, long routes, complex environmental conditions such as terrain and weather, the efficiency of inspections and the accuracy of data have not been high. Use video analysis After functional drones and robots, these conditions can be greatly improved.

In the past, many video analysis in these fields used traditional computer vision to analyze and process videos, but now, AI video analysis is rapidly replacing traditional computer vision, because AI video analysis can extract more features and The accuracy is also higher in many applications.
A new adaptive computing solution

Nowadays, innovation in the electronic field can be said to be endless, especially in the field of AI. Developers often come out with some innovative algorithms, and the aforementioned hardware platforms often cannot meet the needs of new algorithms. Therefore, developers must either modify their algorithms or change the hardware platform. The pace of innovation has to be slowed down.

So is there such a platform? The answer is yes, and that is an adaptive computing platform. According to Xilinx's description, the adaptive computing platform can continuously optimize the hardware according to the application and improve the performance of the overall application. For example, developers can perform hardware acceleration on the entire data channel of the system, including AI models, input data processing, and output data processing. It can be said that every link in the middle can be optimized, thereby greatly improving the efficiency of the overall system.

Moreover, the adaptive computing platform now has a very rich ecosystem. In terms of hardware, there are chips and heterogeneous acceleration hardware platforms, which can provide scalar engines, programmable logic, AI acceleration engines, etc.; in terms of software, different operating systems can be selected, and embedded runtime customized hardware, hardware IP, acceleration Libraries, etc.; in terms of development tools, for FPGA hardware developers, there is a Vivado development platform suitable for low-level hardware implementation, hardware IP, and hardware acceleration library development; for software developers, there are high-level languages ​​that developers are familiar with Vitis integrated development environment for programming in C++, C++ and Python; for AI developers, there is a Vitis AI development platform that can use frameworks familiar to developers (such as TensorFlow, PyTorch, and Caffe frameworks), and developers can use Vitis AI to deploy Their model, and let the model efficiently execute on Xilinx's hardware platform.

In addition, the Xilinx App Store is also part of the adaptive acceleration ecosystem. According to statistics, there are currently more than 200 Xilinx Alveo accelerator cards and cloud-related applications in the application market. Even the Kria SoM adaptive module product launched in July this year has 7 related applications for developers to choose from. Of course, the applications of different workloads are still increasing.

It is worth mentioning that in order to improve developer skills and help developers design better projects on the adaptive computing platform, Xilinx launched a new Xilinx Developer Program last year. The plan includes three aspects:

One is to provide a brand new developer's website, which has all the resources of adaptive computing, including Vivado, Vitis, Vitis AI and other related development resources, as well as some projects, engineering and underlying design.
The second is to provide developers with more than 40 free training modules, including on-demand training courses created based on Xilinx’s computing platform. Developers can enjoy a 50% discount if they join the Xilinx Developer Program.
Third, Xilinx has launched a global design competition, which is currently the most diverse competition in the FPGA developer community.
In addition, Xilinx will also launch related developer certification programs, with different training courses and exams for different developers. After passing, you can get the developer skills officially recognized by Xilinx. Certified developers will get the Early Access program opportunity provided by Xilinx, and get early access to the latest Xilinx technology.
The most important thing is that this developer program is free for all developers. Therefore, any developer can join the Xilinx Developer Program directly through Xilinx’s developer website.
2021 Adaptive Computing Challenge
Now there is an opportunity to quickly enter the AI ​​reasoning market, which is to participate in the 2021 Adaptive Computing Challenge jointly organized by Xilinx and Hackster.io. The core goal of this adaptive computing challenge is to use Xilinx's adaptive computing platform to design with Vitis, Vitis AI or Vivado ML to solve some real-world problems. The majority of developers can quickly master the design skills of the adaptive computing platform through the adaptive computing challenge, and make good use of software and hardware design resources for independent innovation.
 

This year is the second Adaptive Computing Challenge. Compared with last year, this year's scale is larger. The competition will provide a total of 70,000 US dollars in prize money to the 15 winners who finally stand out, as well as more than 350 sets of development boards as hardware support. In addition, the "2021 Adaptive Computing Challenge" has added two special awards, namely the University Program (XUP) and Women in Science and Technology (WIT), aiming to recognize college students and encourage women in science and technology to innovate. Contest registration has been opened during Xilinx Adapt this week.
 

According to the plan, participants of the 2021 Challenge will compete on three major tracks: edge computing, data center AI and big data analysis, including the first to experience a new big data analysis development board for big data and related applications. The highest prizes for the three categories are all US$10,000.
 

In addition, the Xilinx University Program also sponsored the Adaptive Computing Challenge, which will provide an additional $2,500 bonus for college student projects in the three categories to encourage students to innovate in the field of adaptive computing acceleration. More than that, encouraging and praising women's talents is also one of Xilinx's important initiatives. In order to further support the diversification of the technology field, Xilinx has also established a new "women in technology award" with a prize of $2,500. Teams with at least half of the team’s female members are eligible to apply for this additional award while submitting entries for edge computing, data center AI, and big data analysis.
 

To register for the "2021 Adaptive Computing Challenge", please visit the following website: https://www.hackster.io/contests/xilinxadaptivecomputing2021

Or scan the QR code to view immediately
 

According to the schedule, at the Xilinx Adapt Virtual Technology Conference on September 7th, US time, Xilinx will announce that the competition is open for registration. At that time, developers can register to participate in the competition, submit a competition plan and apply for the required hardware platform. Xilinx will conduct a preliminary selection according to the contest plan of the contestants and allocate free hardware. It is expected that the shortlisted teams will be able to get the relevant hardware and start development in mid-November. There is an important date that developers need to pay special attention to, that is, the deadline for competition registration is October 15, 2021. Interested developers must form a team in advance and submit their initial ideas and hardware applications. The competition ends on February 28 next year, when Xilinx will announce the winner of the competition.
Concluding remarks
A brand-new tens of billions-dollar track is coming, and developers can fully release your creativity and imagination to make more and better products. The adaptive computing platform can help you. Join the developer community of the adaptive platform now. Not only can you enjoy various benefits, you can also effectively use the resources of the website and store. At the same time, you can also compete with global developers through the adaptive computing competition. Compete on the same stage. Developers, take action!