Kioxia Corporation today announced the development of a groundbreaking AI-driven image recognition technology designed to automatically identify products moving through logistics workflows. Developed in collaboration with Tsubakimoto Chain Co. (Tsubakimoto Chain) and EAGLYS Inc. (EAGLYS), the new system strengthens automation and operational efficiency across logistics environments. By enabling rapid and accurate product recognition, the technology helps logistics operators adapt quickly to dynamic market conditions, reduce operational burdens, and maintain high service performance despite rising complexity and labor constraints. Central to the new system are KIOXIA’s AiSAQ™ and Memory-Centric AI technologies, which address the industry’s pressing need for scalable artificial intelligence capable of supporting rapidly expanding product categories. The jointly developed solution will debut publicly at the 2025 International Robot Exhibition.
Growing Challenges in Modern Logistics
In recent years, global supply chain and logistics networks have undergone radical changes, largely driven by the continued expansion of e-commerce and the growing diversity of products demanded by consumers. As more transactions move online, logistics facilities are faced with:
- Higher throughput requirements, with increasing quantities of goods entering and exiting distribution centers daily
- Broader product variety, including short-run, seasonal, customized, and limited-edition items
- Greater expectations for accuracy and delivery speed
- Labor shortages, making manual sorting and verification increasingly unsustainable
Traditional methods for identifying and routing goods—whether handled manually or through conventional automation systems—are becoming insufficient. Manual inspection is prone to error and constrained by workforce limitations. Meanwhile, conventional image recognition-based AI solutions require extensive training and periodic updates to maintain accuracy, particularly when new products are introduced.
Limitations of Conventional AI Image Recognition
The majority of existing image recognition systems rely on deep learning models trained on large datasets that represent known products. While accurate under stable conditions, these systems face significant challenges as catalog sizes grow:
- Continuous retraining is required whenever product variations are introduced, consuming time and engineering resources
- Model parameter tuning becomes increasingly complex as data scales
- Computational load increases, driving up power consumption
- Costly and inefficient reprocessing cycles reduce responsiveness when new product lines are released
For logistics operators managing tens of thousands of SKUs that may evolve weekly or seasonally, these limitations can create delays, inefficiencies, and increased operational risk.
A New Approach: KIOXIA AiSAQ™ and Memory-Centric AI
Kioxia’s new system solves these challenges through an innovative architecture built around AiSAQ™ software and Memory-Centric AI technologies. Instead of retraining the entire AI model each time new product information becomes available, the system stores high volumes of real product data directly in high-capacity memory, including:
- Product images
- Label and packaging variations
- Feature descriptors and classification metadata
This flexible storage-based approach enables rapid expansion of recognition capabilities without modifying the base model, significantly reducing downtime and computational demands.
As data volume grows, searching through billions of stored parameters can become slower and more memory-intensive. To overcome this, the new system:
- Indexes memory-stored data
- Transfers indexed data to SSD storage for fast access
- Retrieves relevant data efficiently without degradation in performance
This combination of features enables scalable AI deployment supporting continuously growing product libraries and real-time recognition speeds appropriate for live logistics operations.
Live Demonstration at the 2025 International Robot Exhibition
The jointly developed system will be showcased at the 2025 International Robot Exhibition, taking place December 3–6 at Tokyo Big Sight, where it will be featured in the Tsubakimoto Chain Booth (E6-23). Recognized as one of the world’s largest exhibitions for industrial robots and automation solutions, the event gathers leaders from manufacturing, logistics, robotics, and AI fields to unveil cutting-edge innovations shaping the future of automation.
During the demonstration, visitors will observe the automated image recognition workflow in real time:
- Products move along a conveyor line.
- The system captures visual data as items pass through.
- AiSAQ™ processes the information and retrieves matching data from indexed storage.
- Items are rapidly recognized and classified based on stored labels and product attributes.
Through the demonstration, attendees will see how logistics facilities can improve accuracy, reduce dependency on manual labor, and handle broad SKU variety—including sudden changes in packaging or seasonal goods—without delay or technical interruption.
Driving the Future of Automated Logistics
As businesses continue to accelerate digital transformation, the ability to deploy AI quickly and scale efficiently will be central to logistics competitiveness. Kioxia’s Memory-Centric AI approach offers several major advantages for operators seeking modernization:
- Improved adaptability as product lines change and expand
- Reduced operational cost by eliminating frequent model retraining
- Lower power consumption through optimized memory and storage access
- Higher recognition accuracy and processing efficiency
- Shorter deployment timelines for AI-driven automation tools
This technology positions logistics infrastructure to evolve in step with the demands of modern commerce—advanced, flexible, and capable of supporting large-scale complexity.
Source Link:https://www.businesswire.com/
