Huawei's Open-Source Low-Level Vision Kit Boosts AI Breakthroughs, Expanding Machine Vision Applications
-
Developed by Huawei's Noah's Ark Lab based on the Ascend MindSpore AI framework. According to IT Home on November 21, Huawei's MindSpore Editing has officially released version 1.0 on the Ascend open-source community, making it available to developers.
Review: The MindSpore Editing low-level vision kit, developed by Huawei's Noah's Ark Lab, is built on the Ascend MindSpore AI framework. It implements and unifies SOTA algorithms and models for mainstream low-level vision tasks, providing user-friendly training and inference interfaces. Low-level vision refers to a category of scenarios in computer vision where algorithms process visual signals at the pixel level, enabling tasks such as image and video restoration, quality enhancement, and style transformation.
Breakthroughs in AI technology are expanding the reach of machine vision. On April 5 this year, Meta released the Segment Anything Model (SAM), a major advancement in image segmentation technology. On April 17, Meta introduced the DINOv2 visual model, which achieved impressive results in tasks like semantic segmentation, instance segmentation, depth estimation, and image retrieval—all without requiring fine-tuning due to its self-supervised nature.
CITIC Securities believes that industrial machine vision is currently highly customized, with defect detection being difficult to quantify. Vision models like SAM are expected to increase the feasibility of more projects, opening up new opportunities for adoption. As the industry rapidly develops, companies with strong customer relationships, deep industry experience, and technical expertise are likely to benefit.
Regarding companies, according to disclosures on listed company interaction platforms:
- Juzi Technology: The company's main products include machine vision equipment, control cable assemblies, and control units, supplying clients like Apple and Huawei.
- United Optics: The company is an optical system solutions provider, offering machine vision products and smart display solutions to Huawei.
What is the outlook for the machine vision market?
Machine vision is a rapidly growing branch of artificial intelligence. While definitions vary slightly among authoritative institutions, machine vision essentially involves using machines to replace human eyes for detection, judgment, and control. Developed from a biomimetic perspective, machine vision simulates human vision by capturing images via visual sensors and processing/recognizing them through image processing systems.
Driven by increasing technical demands in industrial automation, the global machine vision industry emerged in the 1960s, matured by the 1990s, and continues to grow rapidly today. China's machine vision industry, which started in the late 1990s, is relatively late compared to the global market but is now in a phase of rapid development.
From the perspective of the industrial chain, the upstream supply of machine vision mainly includes components such as light sources, industrial lenses, industrial cameras, image capture cards, and machine vision software and algorithms. The midstream consists of machine vision system integrators and equipment manufacturers. The downstream applications cover industries like electronics, semiconductors, packaging, food and beverage, automotive manufacturing, robotics, medical devices, and logistics.
Machine Vision Market Status and Competitive Landscape Analysis
Globally, the machine vision market has shown continuous growth from 2015 to 2021. According to GGII data, the global machine vision market reached 80.4 billion yuan in 2021, a 12.15% year-on-year increase, representing a growth of 42.612 billion yuan compared to 2015.
Today, China has become one of the most active regions in the development of machine vision worldwide, with applications spanning industries such as manufacturing, agriculture, pharmaceuticals, military, aerospace, meteorology, astronomy, public security, transportation, safety, and scientific research.
A key reason for this is China's emergence as a global manufacturing hub. The demand for high-precision component processing and advanced production lines has brought many internationally advanced machine vision systems and expertise into the country.
Although machine vision has only been developing for a few decades, the rise of a new wave of global technological revolution and industrial transformation has spurred rapid growth in the industry. Machine vision applications have expanded from their initial use in automotive manufacturing to widespread adoption in consumer electronics, pharmaceuticals, food packaging, and other fields. With economic advancements, 3D machine vision is also gaining traction.
3D machine vision is primarily used for grading products such as fruits and vegetables, wood, cosmetics, baked goods, electronic components, and pharmaceuticals. It enhances the production capacity of qualified products by identifying and discarding substandard items early in the process, reducing waste and saving costs. This functionality is particularly suited for imaging product attributes like height, shape, quantity, and even color.
From the perspective of the industrial lifecycle, the international machine vision industry has reached maturity. Over the next few years, Europe, the U.S., and Japan are expected to continue innovating in machine vision technology, maintaining and even expanding the current market size.
In contrast, China's machine vision industry is still in its growth phase. However, in recent years, the country has accumulated sufficient technology, market experience, and industry expertise, positioning it for rapid development.
Machine vision systems are characterized by their ability to enhance production flexibility and automation. They are commonly used to replace human vision in hazardous work environments or situations where manual inspection is inadequate. In large-scale industrial production, machine vision inspection methods significantly improve efficiency and automation levels compared to manual quality checks, which are often inefficient and less accurate. Additionally, machine vision facilitates information integration, serving as a foundational technology for computer-integrated manufacturing.
The machine vision industrial chain primarily consists of upstream raw materials and components, midstream equipment manufacturing, and downstream end-user industries. In-depth applications of machine vision span multiple segments of this industrial chain.
Taking smartphone manufacturing as an example, machine vision can be applied across various stages, including structural component production, module assembly, final product assembly, and processes involving solder paste and adhesives. For instance, the production of an iPhone requires over 70 sets of machine vision systems. Broadly speaking, machine vision serves numerous downstream industries, such as automotive, 3C electronics, semiconductors, food and beverage, photovoltaics, logistics, pharmaceuticals, printing, glass, metals, and wood.
Statistical data shows that from 2015 to 2021, the number of financing cases and total investment in China's machine vision sector exhibited an overall upward trend. The number of investment events in the industry increased from 42 to 91, while the investment amount surged from RMB 1.049 billion to RMB 19.34 billion.
In terms of segmented application markets, 3C electronics, automotive, and semiconductors (including integrated circuits, PCBs, electronic components, wafers, and flat-panel displays) represent the three major downstream sectors for machine vision in China.
In 2021, the market shares of these segments in China's machine vision industry were 31.56% for 3C electronics, 11.04% for automotive, and 10.22% for semiconductors. Other notable sectors included lithium batteries (8.85%), pharmaceuticals (8.05%), and food packaging (6.29%).
Future Trends in the Machine Vision Industry
With increasing R&D investments and technological advancements by domestic machine vision companies in recent years, China's reliance on imported machine vision solutions has gradually decreased.
In 2021, domestic brands accounted for 58.43% of the Chinese machine vision market, while foreign brands held 41.57%. As the localization process accelerates, the market share of domestic machine vision companies is expected to continue growing in the future.
The trend toward miniaturization in the machine vision industry allows more components to be packed into smaller spaces, meaning machine vision products are becoming more compact. This enables their application in the limited spaces available within factory premises.
For example, LEDs have become the dominant light source in industrial components. Their small size simplifies the determination of imaging parameters, while their durability and stability make them highly suitable for factory equipment.
The realization of more functionalities primarily stems from enhanced computing power, higher-resolution sensors, faster scanning rates, and improved software capabilities. As PC processor speeds steadily increase, their prices are simultaneously declining. This has spurred the emergence of faster buses, which in turn allow larger images with more data to be transmitted and processed at higher speeds.
Machine Vision Industry's Future Development Plans and Prospects
In January 2022, China's State Council issued the 14th Five-Year Plan for Digital Economy Development, proposing the intelligent transformation of agricultural, forestry, animal husbandry, and fishery infrastructure and production equipment, along with the application of technologies such as machine vision and machine learning.
In August 2022, the Ministry of Science and Technology and other departments jointly released the Guidelines on Accelerating Scenario Innovation to Promote High-Level AI Applications for High-Quality Economic Development. The document highlights the priority exploration of intelligent scenarios in manufacturing, including industrial brain systems, robotic-assisted manufacturing, machine vision-based industrial inspection, and equipment interconnectivity management.