Smart Vision-Based Inventory System for Real-Time Sack Monitoring and Predictive Stock Optimization

Abstract

The present study aims to implement an intelligent system for rice sack inventory management that integrates object recognition, object counting, and object forecasting. Existing solutions for rice inventory management are characterized by manual object counting processes that lack real-time data and result in poor transparency, ultimately negatively impacting the supply chain. To address these, the proponents of the study developed and implemented an integrated Vision-Based Inventory Intelligence System using YOLOv11 for object recognition and counting, as well as ARIMA for object forecasting. The artificial intelligence features of the web system were trained on a dataset of 1,000 images. The result attains a mAP of 98.07% with a Mean Absolute Percentage Error (MAPE) <10% for object forecasting. The system's architecture consists of a high-speed camera, a relational database, and an automated dashboard that updates in less than 5 seconds. The system is built using the Design Science Research methodology, in accordance with ISO/IEC 25010 standards, and SDGs 9, 12, and 16. The system aims to improve warehouse management through intelligent automation.