Implementation of Deep Learning CNN Algorithm for Classification of Gas Station Digitization Inventory Devices

Abstract

The digitization of inventory management in gas stations has become increasingly crucial for improving record accuracy, operational efficiency, and the reliability of device monitoring. However, the identification of inventory devices is still widely performed manually, making the process vulnerable to recording errors, data inconsistencies, and delays in audit activities. This study aims to implement a Deep Learning–based Convolutional Neural Network (CNN) algorithm to automatically classify inventory digitization devices in gas stations, including digital sensors, fuel pumps, control panels, electronic dispensers, and monitoring modules. The image dataset of these devices was collected through direct acquisition and further expanded using augmentation techniques to address limited data availability. The CNN method was developed through several key stages, including image preprocessing, automated feature extraction through convolutional layers, and model training using regularization and hyperparameter tuning to achieve optimal performance. The model’s performance was evaluated using accuracy, precision, recall, F1-score, and a confusion matrix. The results indicate that the CNN model achieved a classification accuracy of XX% (to be updated after experimentation), demonstrating its capability to effectively recognize visual patterns of inventory devices. These findings confirm that deep learning provides a reliable automation solution for gas station inventory digitization processes and contributes to more efficient asset management and data-driven decision-making