Imagine a world where logistics operate with unparalleled speed and accuracy, effortlessly adapting to the ever-changing flow of goods. But what if current AI systems can't quite keep up with the increasing complexity? Kioxia Corporation, in collaboration with Tsubakimoto Chain Co. and EAGLYS Inc., is tackling this challenge head-on with a groundbreaking AI-driven image recognition technology designed to revolutionize logistics processes. This isn't just about incremental improvements; it's about enabling a fundamental shift toward truly automated and efficient operations.
This new system enables automatic identification of products as they move through the logistics workflow. The core idea is to empower businesses to react swiftly to dynamic market demands while keeping costs in check and upholding superior service quality. Kioxia's AiSAQ™ and Memory-Centric AI technologies are the cornerstone of this innovation, specifically addressing the growing need for AI solutions that can scale effectively as product ranges become increasingly diverse. Think of it like this: as online shopping booms and the variety of products explodes, the AI needs to handle a much larger and more complex 'visual vocabulary.'
But here's where it gets controversial... Traditional AI image recognition relies on deep learning models. These models often require extensive parameter tuning and retraining every time new or seasonal products are introduced. This retraining process can be incredibly time-consuming and resource-intensive, leading to higher power consumption and increased operational costs. This is especially true when dealing with massive product catalogs, a common scenario in today's e-commerce driven world. Some argue that these costs are simply the price of progress, but Kioxia believes there's a better way.
The Kioxia AiSAQ software, in conjunction with Kioxia's Memory-Centric AI technology, offers a compelling alternative. This approach leverages high-capacity storage to house vast amounts of product data, including images, labels, and feature information. This allows for the rapid addition of new product information without requiring a full retraining of the base AI model. It's like having a constantly updated visual encyclopedia for the AI to reference. And this is the part most people miss... To combat the potential for longer search times and increased memory demands as the data volume grows, the technology employs a clever indexing system. This system intelligently organizes the memory-stored data and strategically moves indexed data to SSD storage, facilitating faster and more efficient information retrieval. This ensures speedy and accurate product identification, even with a massive and ever-changing product database.
This collaborative technological achievement from Kioxia, Tsubakimoto Chain, and EAGLYS will be showcased at the upcoming 2025 International Robot Exhibition, taking place December 3–6 at Tokyo Big Sight (Tsubakimoto Chain Booth E6-23). The International Robot Exhibition is a leading global event highlighting cutting-edge automation advancements in manufacturing and logistics. Visitors to the booth will have the opportunity to witness the jointly developed image recognition system in action. The demonstration will illustrate how the system captures image data of products moving along a conveyor belt and rapidly classifies them by referencing stored features and labels. This provides a tangible example of how logistics operations can manage a wide and constantly evolving range of products with enhanced accuracy and efficiency.
For those eager to explore the software, KIOXIA AiSAQ open-source software is available for download at: https://github.com/kioxia-jp/aisaq-diskann
Notes:
1: More information on KIOXIA AiSAQ Technology Designed to Reduce DRAM Requirements in Generative AI Systems Released as Open Source Software can be found here: https://www.kioxia.com/en-jp/business/news/2025/20250128-1.html
2: Further details on the Development of Image Classification System Deploying Memory-Centric AI with High-capacity Storage are available here: https://www.kioxia.com/en-jp/rd/technology/topics/topics-39.html
*KIOXIA AiSAQ is a trademark of KIOXIA.
All other company names, product names and service names may be trademarks of third-party companies.
About Kioxia
Kioxia is a global leader in memory solutions, committed to the development, production, and sale of flash memory and solid-state drives (SSDs). Kioxia's journey began in April 2017, when it was spun off from Toshiba Corporation, the pioneering company that invented NAND flash memory in 1987. Kioxia's mission is to enrich the world with "memory" by providing products, services, and systems that empower customers and create memory-based value for society. The company's innovative 3D flash memory technology, BiCS FLASH™, is shaping the future of storage in a wide range of high-density applications, including advanced smartphones, PCs, automotive systems, data centers, and generative AI systems.
Note: Information presented in this document, including product prices and specifications, service details, and contact information, is accurate as of the date of the announcement but is subject to change without prior notice.
So, what do you think? Is this memory-centric approach the key to unlocking the next level of AI-powered logistics, or are there limitations that haven't been fully addressed? Share your thoughts and opinions in the comments below!
