Cultivating 20 billion-yuan partners! Huawei ignites new strategies and new products in the machine vision field

Enthusiast Zhang Ying

Recently, the track has heated up, and there are mainly two good news. First, on August 8, the machine vision army was officially announced, and Huawei's machine vision partner operation was officially launched. In the future, more than 20 billion yuan partners will be cultivated. Huawei's executive chairman Wang Tao said that it will increase investment in image, view coding, visual intelligence, etc., and build end-edge, cloud full-scene products with camera definition as the core.

In May, high-end equipment and machine vision Cosinex completed nearly 400 million yuan in Series C financing. This round of financing was jointly led by SMIC Juyuan, Minsheng Equity Investment Fund, and Hongtai Fund, and followed by industrial investors such as Xinrui Capital and Share Investment. Yunxiu Capital served as the exclusive financial advisor for this round of financing.

What is the market trend of the machine vision industry in the next three years? What is Huawei's machine vision strategy? What are the highlights of the new products? This article conducts a detailed analysis.

Reach 35.6 billion in 2025! The compound growth rate of the machine vision market in the next three years will reach more than 19%

According to data from Markets andMarkets, the global machine vision device market size maintains a steady growth trend and is expected to reach US$14.7 billion in 2025. In the future, the penetration rate of machine vision in traditional manufacturing industries will increase, and the market demand generated by emerging industries (security monitoring, logistics sorting, and smart transportation) is expected to drive the global machine vision industry to continue to grow steadily.

According to data from the China Machine Vision Industry Alliance, the scale of China's machine vision industry has grown from 10.18 billion yuan in 2018 to 14.42 billion yuan in 2020, with a compound growth rate of 19.02%. It is expected to reach 29.60 billion yuan in 2023. CICC pointed out that China's machine vision market has broad application scenarios, but this market has low penetration and weak levels in some scenarios, and is in its early stages. It is expected that with the increase in penetration and localization rates in the future, the market space for machine vision in which domestic manufacturers in China participate is expected to reach 35.6 billion in 2025.

Huawei Machine Vision Company is one of the important participants in the machine vision market. How to quickly tap into the important development opportunities in this track? The strategy announced at the Huawei Machine Vision Corps press conference on August 8 is worth pondering.

Huawei's machine vision strategy upgrade! Covering four major aspects from products to ecology

"In 2021, Huawei opened a new development direction for the industry with the collaboration of terminals and clouds. In 2022, the machine vision product line was upgraded to the Legion. In the past four years, machine vision products and ecology have become increasingly rich and reached a new height. Whether in terms of technology, algorithm application, or industry standards, Huawei Machine Vision and its partners have achieved great success in recent years. Huawei uses its own strength to empower the entire security concept and build an open, symbiotic, intelligent and win-win ecological system. This is the foundation for the future development of the Machine Vision Legion." Duan Aiguo, CTO & COO of Huawei Machine Vision Legion and President of Products and Solutions, said.

Hong Fangming, CEO of Huawei Machine Vision Legion, pointed out that the strategic upgrade of Huawei Machine Vision Legion has begun in an all-round way, mainly focusing on four major aspects: First, in terms of market strategy, the short-chain operation of the Machine Vision Legion will be closer to the market. Second, in terms of product capabilities, the Legion will create more competitive products, build a more open product system, and improve the experience, marketability and differentiation of the full range of cloud-edge products.

Third, in terms of services, the Legion will create a service capability center and build a unified and efficient service system. Hong Fangming has traveled to more than 20 provinces to meet with customers in the past two months to discuss the improvement of interactive service experience. Fourth, in terms of ecology, the Legion will build a more open ecological system. The traditional security industry is severely fragmented, with many data islands, and many partners are unable to participate in the project; in the era of machine vision, the Legion will build an open symbiotic, intelligent and win-win ecological system.

Figure 1: Huawei's machine vision army strategic upgrade Electronics enthusiast screenshot

Summarizing the achievements of the machine vision product line in the past three years, Duan Aiguo believes that three things need to be done right: First, inclusive AI. Huawei machine vision aims to develop artificial intelligence. Good Hope is synonymous with artificial intelligence. Huawei focuses all its R&D efforts on AI products, and continues to make AI products more universal, cheap and easy to use.

Second, adhere to open symbiosis. In the field of security, many giants have established many advantages through more than 10 years of hard work. The future competitive strategy of Huawei's machine vision army will insist on using openness to fight against closure. Whether it is the architecture of the end-edge cloud or the construction of the industrial alliance, including the launch of regional capabilities; Third, always adhere to mechanism innovation. Whether it is the software-defined architecture or the three major root technologies of SuperColor, SuperCode and AI breakthroughs released last year, Huawei has launched a series of innovative products to provide more innovative products for the industry, customers and channels.

Cracking the pain points of AI technology to enable scene applications Huawei launches the Good Hope AI framework and flagship products

Many AI companies are struggling on the edge of profit and loss. Why does the maturity of AI technology not mean the success of large-scale commercial use? Duan Aiguo analyzed that there are four major problems: First, the core is the algorithm problem. AI must first use algorithms to solve practical difficulties. The specific application scenarios vary greatly, but there are few algorithms available on the market. The pain point is that it is difficult to iterate after the algorithm is loaded, and it is solidified. Taking the high-altitude throwing of smart cameras as an example, Huawei's high-altitude throwing algorithm has been iterated more than 10 times in the first half of this year alone. With such a fast iteration, if Huawei's Shengsi architecture cannot support the rapid evolution of the algorithm, once it comes to commercial application, it will be difficult to enjoy the results of technology iteration. This is an algorithm problem.

Second, the pain point of weak levels in some scenarios. It is reported that the proportion of AI cameras in the security industry does not exceed 15%, because in many application scenarios, channel partners still have no AI products available when facing user needs. Inclusive AI products are not rich.

Third, the project is complex and the cost is high. It only takes 5 minutes to install a traditional camera, and it may take a day to install a smart camera. If we don't work hard, use the attitude of excellence and the spirit of craftsmen to polish AI products, the landing of AI can only stay in PPT.

Fourth, it is difficult to combine technologies across. The traditional box is a one-dimensional world. Once we get to AI, we need to consider AI algorithms and add another dimension. We also need the business closed loop of AI algorithms and surrounding applications to add another dimension. From one dimension to three dimensions, it is a complex system engineering, involving end-edge cloud network, perception, connection, computing, storage, etc.

Is the current pain point difficult to adapt AI algorithms to hardware? Huawei launched the Good Hope AI framework this year. This framework is a standard, open, and collaborative full-scenario AI architecture built by Huawei machine vision around end-edge cloud products. This architecture

consists of four parts: 1. Good Hope AI model conversion tool, 2. Good Hope AI development framework, 3. Good Hope AI deployment framework, and 4. Good Hope AI reasoning framework. Based on Huawei Good Hope's understanding of machine vision, end-edge cloud full scenarios, and its own software-defined products, etc., the needs of the entire industry and Huawei technology should be strung together to create an end. For example, based on Huawei's Good Hope AI reasoning framework, it can be compatible with the differentiation of chips of different devices, and based on the understanding of the underlying capabilities of Huawei's self-developed chips, chip API operators and other integrations are integrated.

Figure 2: Screenshot of Huawei's Hope AI architecture from an electronics enthusiast

At the press conference, Huawei's machine vision army focused on scenarios and launched five series of new products, including the AI ​​perimeter FLIV series, low-light all-rounder, high-altitude throwing series, smart parking special, a new generation of Good View ball camera and other products.

Figure 3: Screenshot of Huawei FLIV series electronics enthusiast

Duan Aiguo introduced that Huawei Machine Vision has launched the FLIV series this time, with an overall detection rate of up to 98% and an overall false detection rate of less than 2%. It has integrated Huawei's capabilities, with a sound pickup distance of 5 meters, 30 meters of white light fill light, and 50 meters of fill light. The price is greatly reduced compared to previous products.

Figure 4: Screenshot of Huawei's "Mozi 2" camera taken by an electronics enthusiast

Duan Aiguo emphasized, "The 'Mozi 2' of the low-light full-color series is the world's first F0.9 lens camera released by Huawei. This technology comes from Huawei's 2012 laboratory, which introduces the high-precision assembly technology of the terminal industry chain and the high-temperature and long-term stability technology of the automotive industry chain. The current lens used in this product is 4 million pixels, and it has the ability to evolve to 20 million pixels in the future. It has instantaneous focusing technology, which means that the image is still very clear during zooming."

In response to objects thrown from high places, Huawei has launched three products, including M2141 and M2150, which can detect 16-50-story buildings. Duan Aiguo said that the relevant algorithms are developed in cooperation with the industry's high-altitude object throwing algorithm partners, including Intelinda, Bostin, Xiaoshi Technology and other partners are all excellent.

Figure 5: Screenshot of Huawei M2281-QL-T smart camera from Electronics Enthusiasts

For roadside parking, Huawei has released the M2281-QL-T smart camera. This product has a built-in super NPU chip with a computing power of 2Tops. It supports multiple scene algorithms in parallel and can extract multiple structured information such as faces, bodies, non-motor vehicles and vehicles at the same time. It is suitable for intelligent services in urban back streets and alleys, community entrances and exits, and county, township and village roads. The camera has a built-in T-shot patented AI adaptive capture algorithm, which increases the simultaneous capture rate of people/cars at night by 50%, completely solving the problem of multiple targets such as people/cars being unable to be captured at the same time at night. This article is original by Electronic Enthusiasts. Please indicate the above source for reprinting. WeChat ID: zy1052625525. If you need to join the group, please add WeChat elecfans999. If you need to submit news and interviews, please send an email to huangjingjing@elecfans.com.