Build an AI industrial vision large model to empower intelligent manufacturing. "Geyuan Technology" has obtained a $46 million Series B financing | First published on 36Kr.
Text | Wang Fangyu
Editor | Su Jianxun
36Kr has learned that recently, Shenzhen Geyuan Technology Co., Ltd. (hereinafter referred to as "Geyuan Technology") has completed a $46 million Series B financing round. This round of financing is led by UP Partners, and the funds raised will be mainly used to expand the market and increase technological research and development.
Geyuan Technology was established in 2018 and is a company with AI large model technology as its core. Its team members come from Stanford University, Massachusetts Institute of Technology, as well as well-known enterprises such as Google, Bosch, Huawei, and Tencent.
In terms of the company's business, Geyuan Technology focuses on solving the problem of surface defect detection through deep learning technology, helping manufacturing enterprises reduce the human resource cost of quality inspection and significantly improve the quality and yield of products.
Appearance defect detection is an essential process in the manufacturing industry and an "important checkpoint" for identifying defective products and ensuring product quality. With the increase in the precision of industrial manufacturing and the continuous rise in human resource costs, product quality inspection that solely relies on manpower is becoming a bottleneck for manufacturing enterprises to develop towards refinement. More and more enterprises are turning to defect detection systems based on machine vision.
"For detecting a car part, the missed detection rate of AI is 10 times lower than that of manual visual inspectors." Long Defeng, the general manager of Geyuan Technology, told 36Kr. Under the background of reducing costs and increasing efficiency, the currently best domestic production factories are trying every means to apply AI to all aspects and corners of production to improve efficiency and profitability. The demand for the deployment of AI quality inspection applications is constantly increasing.
On the one hand, AI quality inspection can improve the efficiency of enterprise defect detection and save manufacturing enterprise costs. On the other hand, it also avoids missed detections caused by human errors and achieves more efficient and accurate quality control.
Currently, there are three major difficulties in the surface defect detection system of manufacturing products. Firstly, it is difficult to identify random defect forms. Secondly, it is difficult to illuminate the surface of complex materials. Thirdly, it is difficult to image complex geometric shapes.
Detection equipment of Geyuan Technology Image source: Authorized by the enterprise
For this reason, Geyuan Technology has developed two core product modules: the CorteX deep learning module and the OptiX dynamic optical module, to purposefully solve the above three pain points, and can provide a complete set of solutions from AI products only used for defect detection to automated detection.
Among them, the CorteX deep learning module can manage the defect classification level threshold, significantly reducing the misjudgment rate. It is introduced that the CorteX module only requires 5 defect samples for model training, significantly reducing the sample quantity requirement. And in identifying complex defect forms, the missed judgment rate of the CorteX module is 10 times lower than that of human visual inspection.
The OptiX lighting imaging system can achieve 2³² different lighting methods, fully adapting to the lighting requirements of different types of defects, quickly adjusting and adapting to environmental changes and new defects. At the same time, it also supports a flying shooting speed of 0.5m/s, significantly improving the online detection speed.
It is introduced that with the cooperation of the two core product modules, a detection device of Geyuan Technology can replace 4 manual visual inspectors, and the missed detection rate can be reduced by 10 times compared to manpower.
Up to now, Geyuan Technology has more than 100 cooperative customers. The team's independently developed intelligent imaging system OptiX and deep learning system CorteX have formed a market leading advantage and entered the supply chain of many global leading lithium battery, consumer electronics, new energy electric vehicle and auto parts manufacturing enterprises.
According to Long Defeng, the general manager of Geyuan Technology, currently, Geyuan's downstream customers cover 10 of the top 30 lithium battery manufacturers in the world, 9 of the top 100 global leading consumer electronics 3C manufacturing enterprises, 10 of the top 50 global leading auto parts manufacturing enterprises, and 2 of the top 4 global new energy electric vehicle manufacturing enterprises. "The world's leading enterprise customers will collect all the pictures and judgment data generated by the visual systems we provide as valuable data assets for improving the process," Long Defeng told 36Kr.
At present, China's economy has shifted from a stage of high-speed growth to a stage of high-quality development, urgently requiring major innovative technologies such as a new generation of artificial intelligence to add fuel and continue to drive. It has become a general trend for AI technology to empower intelligent manufacturing.
"The goal of Geyuan Technology is to become the best eyes and brain for factories." Long Defeng told 36Kr. In the next step, Geyuan will continue to deepen research and development innovation, promote the landing and application of large models in the production field, and enable AI technology to better empower the manufacturing industry.