Computer Vision-Enabled Inventory Management System: A Cloud-Native Solution for Retail Cost Reduction

In order to reduce Stock Keeping Unit (SKU) storage redundancy and to build a computer vision-based inventory management system, this research is based on an AWS cloud-native architecture, integrating the image recognition capability of the YOLOv7-B6 model and the elastic scheduling mechanism of AWS Lambda and AWS Fargate to optimize the compression and heterogeneous deployment scheduling of SKU-level data. The average compression ratio of the analyzed WebP and Zstandard dual-layer coding structure on 1920×1080 resolution images reaches 0.264, the redundancy rejection rate is improved to 37%, and the volume of a single SKU image is reduced from 2.8MB to 0.74MB, which reduces the storage cost by 40.12%. The average service latency of the AWS-based system is controlled at 85ms under 50 concurrent requests, and the load balancing ratio is optimized from 1.52 to 1.08. The results show that the inventory management system based on the fusion of visual recognition and containerized scheduling can effectively improve the efficiency of inventory flow and achieve structural cost reduction.