KIOXIA Europe has unveiled new AI-driven image recognition technology designed to automatically identify products in logistics workflows. Developed in collaboration with Tsubakimoto Chain Co. and EAGLYS Inc., the system aims to boost automation, streamline operations and help logistics providers respond more effectively to changing market conditions—while keeping costs under control and maintaining high service quality. Central to the new recognition platform are KIOXIA’s AiSAQ[1] and Memory-Centric AI[2] technologies, which together enable scalable AI deployment in logistics environments. The technology will be showcased at the International Robot Exhibition 2025 in Tokyo.

As e-commerce volumes rise, logistics networks are processing increasing quantities and a growing variety of goods. At the same time, a persistent shortage of skilled workers is intensifying the demand for more efficient operations—an area where AI offers clear advantages. Yet conventional AI image recognition still depends on deep learning models that need regular parameter tuning and retraining whenever companies add new or seasonal products. This not only delays processes but also drives up energy use and operating costs, making it particularly inefficient for large product catalogues.

KIOXIA’s AiSAQ software, combined with its memory-centric AI approach, is designed to eliminate these bottlenecks. The system stores large volumes of new product data—images, labels and product details—in high-capacity memory, allowing companies to integrate new items quickly without retraining the core model.

To counter potential slowdowns and rising storage demands as data volumes grow, the system indexes the memory-stored data and transfers the indexed sets to SSD storage. This ensures faster, more efficient data retrieval even as catalogues expand.

“Our goal at KIOXIA is not only to provide the best storage options for applications, but also to offer support and make our technology accessible. That’s why we’re making it available as open source. In this way, we’re helping developers and system architects optimise performance and capacity in innovative new ways,” says Axel Störmann, Vice President and Chief Technology Officer for Memory and SSD Products at KIOXIA Europe. “By using SSD-based ANNS, we reduce dependence on costly DRAM while meeting the performance requirements of leading in-memory solutions. This allows us to expand the performance range of extensive RAG applications.”

KIOXIA and EAGLYS will present the result of their collaboration at the International Robot Exhibition 2025, held in Tokyo from 3 to 6 December. As one of the world’s leading events for automation in manufacturing and logistics, the exhibition offers visitors a frontline view of emerging technologies. At the Tsubakimoto Chain Co. booth (E6-23), attendees can see the image recognition system in live operation: as products move along a conveyor belt, the system captures image data and classifies items according to stored attributes and labels. The demonstration highlights how logistics companies can manage diverse and fast-changing product ranges with greater precision and efficiency.

The open-source software KIOXIA AiSAQ is available for download at: https://github.com/kioxia-jp/aisaq-diskann

Notes
[1] Announcement of the release of KIOXIA AiSAQ technology to reduce DRAM requirements in GenAI systems as open-source software: https://www.kioxia.com/en-jp/business/news/2025/20250128-1.html
[2] Background article on the development of an image classification system based on memory-centric AI and high-capacity memory: https://www.kioxia.com/en-jp/rd/technology/topics/topics-39.html