Kioxia America, Inc. has announced a collaborative project with Kioxia Corporation, Tsubakimoto Chain Co., and EAGLYS Inc. to develop AI-driven image recognition technology for logistics workflows. The system is designed to automatically identify products moving through warehouses and distribution networks, improving efficiency, cutting costs, and helping companies respond quickly to changing market demands.

Central to the project are KIOXIA AiSAQ™ and Memory-Centric AI technologies, which allow for scalable AI adoption as product types diversify. The jointly developed system will be demonstrated at the 2025 International Robot Exhibition in Tokyo from 3–6 December.

The rise of e-commerce has increased the volume and variety of products handled by logistics networks. At the same time, ongoing labour shortages are driving demand for automated solutions. Traditional image recognition AI systems require retraining and parameter tuning whenever new or seasonal products are introduced. This process is time-consuming and can raise both operational costs and energy consumption, particularly when managing large product catalogues.

KIOXIA AiSAQ software, paired with Memory-Centric AI, addresses these issues by storing extensive new product data, including images, labels, and feature information, in high-capacity storage. This allows new products to be added without retraining the base model. To maintain performance as data volumes grow, the system indexes memory-stored data and moves it into SSD storage, enabling faster retrieval through Retrieval Augmented Generation (RAG), a form of Memory-Centric AI used in image recognition.

At the International Robot Exhibition, visitors will see the system in action at Tsubakimoto Chain Booth E6-23. As products move along a conveyor, the technology captures images and rapidly classifies items by referencing stored labels and features. The demonstration highlights how logistics sites can manage a continually changing range of products with improved accuracy and speed.