Introduction of AI/IVA use cases for the manufacturing industry and new Jetson products

NVIDIA is widely known around the world as a GPU manufacturer specializing in image processing such as 3D graphics, but we are expanding our offering of platforms combined with services in addition to GPUs in various fields. In this lecture, AI use cases in the manufacturing and industrial fields were introduced, and Mr. Umemoto said that there are five major uses of AI in these fields.

Among them, we picked up and introduced specific AI introduction cases from "optical visual inspection" and "IVA (Intelligent Video Analytics) video analysis".

・ AI camera using GPU detects cracks in sheet metal in the press process in a few seconds

・Inspects micron-level defects in electronic devices, realizing inspection accuracy that exceeds the level of humans

・ AI-equipped drones acquire image data in real time, streamlining maintenance work

・Improve the safety of the site by detecting people even from a special angle in a dangerous site

Mr. Umemoto said that as the use of AI spreads, there are three challenges for developers in the AI world (see image). However, since these issues can be resolved by using the most suitable devices, tools, and services, he concluded by recommending that you consider using NVIDIA and its partners.

Achieve high-precision real-time behavior analysis with a small amount of data! Examples of use for improving on-site work and preventing accidents

In general, in order to introduce AI, it is first necessary to acquire a huge amount of data and use it to make AI learn. However, this hurdle is so high that many people are reluctant to take the plunge.

"VP-Motion", introduced by Mr. Fujita, Board Director of Next System Co., Ltd., makes it possible to detect work (behavior analysis) simply by reading a few videos of actual work. In this lecture, Mr. Fujita actually showed a real-time skeleton detection demo.

This "VP-Motion" consists of a "learning part" and a "detection part". It is executed and behavior analysis in AI can be realized.

Until now, AI could not be realized without a GPU installed in a high-spec PC. However, by utilizing NVIDIA 's latest product Jetson AGX Orin, palm-sized edge computing used in this demonstration, real-time image analysis with server-class processing performance becomes possible. The Jetson AGX Orin version of "VP-Motion" will be released in mid-August.

The Forefront of AI Visual Inspection - Inspection Circumstances Unique to the Food Industry and Full Automation through Cooperation with Robots and Ejection Mechanisms -

​YE Digital Co., Ltd. has been building systems mainly for the manufacturing industry, but recently Mr. Tsubasa of the company says that they are focusing on food manufacturing sites in particular. On the other hand, there are issues such as labor shortages at manufacturing sites, the need to maintain excessively high quality, and variations in judgment standards among workers. He introduced "MMEye", a Visual inspection automation solution that uses AI to solve such problems.
　

In the case of food, it was not possible to maintain high inspection accuracy with the conventional rule-based system because the determination of good and bad products is ambiguous. Therefore, by utilizing AI deep learning, automatic recognition can be performed with higher accuracy than humans, and it can strongly support automation in the field. There are still many places where final checks are done by people, but by introducing such a solution, it is possible to improve quality, reduce costs, and save labor.

In the second half, Mr. Sakumoto from iCube Digital Co., Ltd., which was jointly established by Yaskawa Electric and YE Digital, spoke. Taking advantage of the company's strength, "there are many employees with know-how not only in the IT field but also in the edge field, and we can work together", we have developed the "MMEye" introduced by Mr. Murasaki and a robot-based discharge mechanism. He explained an example of automating the entire inspection process by linking.

The "Y's-Eye Hamburger", a model specialized for inspection, which was released in July​ ​2022, is a set of flows from visual inspection by AI to exclusion mechanism to analysis. By limiting the target to frozen hamburger patties, it is possible to specify the customization expected for each customer to some extent, and as a result, we have achieved introduction in a shorter period of time than before.

As shown in the diagram above, it is possible to analyze the trend of defective products, implement improvements in the previous process, and create a cycle that suppresses the occurrence of defective products.

Based on these, at the end of the lecture, the points to realize full automation are 1. Utilization of highly accurate AI judgment model, 2. 3. Hardware compatible with high-speed image processing (Jetson AGX Orin, etc.); I was told that there are three groups of FA engineers with OT knowledge.

Breaking through barriers to introducing AI in the manufacturing industry ~ Solving issues with edge devices and image analysis ~

There are various barriers to promoting the introduction of AI within a company​. Many people have one.

Macnica would like to propose a breakthrough to these issues.

・Selection of optimal edge computing as an AI engine and support utilizing technical knowledge

・Introduction support leveraging partnerships with algorithm vendors that have strengths in AI​

the Company have a dedicated AI development team and have accumulated a track record of providing a wide range of support. In this way, we can provide total support for the environment required for AI introduction, so please feel free to contact us.