AI Is Automating Defect Inspection

The Innodisk EXMU-X261 FPGA machine vision solution kit. (Source: Innodisk.)

Taiwan-based Innodisk launched a new machine vision platform this week called the EXMU-X261. The field programmable gate array (FPGA)-based platform targets automated defect inspection, with Innodisk claiming that it can perform inspections faster than a human inspector. Innodisk believes the EXMU-X261 will be in high demand among its enterprise, industrial, and aerospace partners.

The EXMU-X261 is powered by AMD’s Xilinx Kria K26 system-on-module (SoM), which was originally designed to enable smart city and smart factory applications. The EXMU-X261 comes with HDMI 1.4 video output, 1GbE LAN, four USB 3.1 Gen1 ports for cameras and other peripherals, two M.2 slots, and a terminal block for expansion. Innodisk touts the EXMU-X261’s ability to accelerate defect inspection as a way to make the manufacturing process faster, cheaper and free of human error.

The EXMU-X261 supports Innodisk’s AI Suite software development kit (SDK) for deploying and managing AI models.

Innodisk says the EXMU-X261 is built to withstand harsh industrial environments. It operates between 0 and 70 °C and can be customized with industrial temperature support from ˗40 to 85 °C. Manufacturers can also purchase Innodisk’s InnoAgent, an out-of-band remote management module, to remotely manage the EXMU-X261 from off-site locations—even in the event of a system crash or an in-band network failure, according to Innodisk. The company says that InnoAgent allows manufacturers to operate the EXMU-X261 platform unmanned to further reduce labor and maintenance costs.

In a May 2022 press release, Innodisk Chairman Randy Chien said that the fundamental reason why Innodisk is focusing on edge AI is to reduce cloud processing costs and the pressures of data transmission.

Innodisk plans to continue expanding into the edge AI market, with a focus on absolute integration, application insight, and AI empowerment. Johnny Wu, Director of Innodisk’s Intelligent Peripheral Application division, noted that all AI development derives from the end user’s application requirements.

“When the application calls for it, it is necessary to collect, store and utilize high-speed and efficient data,” Wu said in an August 2022 press release. “Doing so lays out an important foundation for artificial intelligence. Ultimately, the road to further analyzing and deducing huge amounts of data to achieve truly intelligent AI applications depends on the support of computing technology.”

The EXMU-X261 follows Innodisk’s July 2022 launch of edge AI solid state drives (SSDs). Innodisk says that Edge AI SSDs help process data at high speeds at the source, which improves latency and lowers costs. As Innodisk continues to offer edge servers and edge data centers, it is measuring how swiftly data utilization shifts from cloud to edge applications.