Warehouse Technology Revolution: Integration of Drones and Deep Learning Algorithms for Stock Identification and Calculation Automation

Abstract

The technological revolution in warehouse management is accelerating, especially with the use of drones and deep learning algorithms to overcome operational challenges. This research aims to develop a deep learning-based automation system that integrates drones for accurate and efficient identification and inventory calculation in retail warehouses. The research method uses the Design Science Research Methodology (DSRM) approach, which includes problem identification, solution development, demonstration, evaluation, and communication. The dataset contains more than 4,000 images of annotated goods and is used to train YOLO, Mask R-CNN, and RetinaNet models. The evaluation was conducted using Precision, Recall, and Average Precision (mAP) metrics. The test results show that YOLO has the best performance with an average mAP of 0.5 of 0.978 and a processing time of only 4.82 seconds, far superior to Mask R-CNN and RetinaNet. In addition, drone integration allows for efficient imaging of goods, reduces the risk of work accidents, and improves inventory accuracy by up to 100%. The results of this study highlight the importantance of combining drone technology and deep learning algorithms to create reliable solutions in inventory management, especially in dynamic retail environments. This research makes a significant contribution to the optimization of logistics processes and warehouse operations, as well as being the basis for further development in the industry.