International Journal of Computer Theory and Engineering

Abstract—Household appliances contribute substantially to residential energy demand, yet appliance-level monitoring remains difficult due to the cost and limited availability of sensing infrastructure. This study presents a systematic comparative evaluation of deep transfer learning models for visual recognition of household appliance positioning, which is a foundational component for future energy-aware systems. A mixed-source dataset of 1500 images (1100 online and 400 real-world field samples) across 11 appliance categories was constructed. Three architectures, DenseNet121, ResNet50, and EfficientNet-B0, were evaluated under identical experimental settings. To ensure robustness and reproducibility, performance was assessed using accuracy, precision, recall, F1-score, an external real-world test set, and 5-fold stratified cross-validation with results reported as mean ± standard deviation. Experimental results show that DenseNet121 achieves the best performance, with 97.0% accuracy, 96.0% validation accuracy, and an F1-score of 98.4%, outperforming EfficientNet-B0 and ResNet50. Cross-validation yields a mean accuracy of 97.0% ± 0.29%, indicating strong stability and generalization on medium-scale datasets. In addition, a prototype application demonstrates how visual recognition can be integrated into a user-facing system that provides contextual energy-related information to enhance user awareness. Rather than directly measuring or predicting electrical consumption, this work establishes a robust visual recognition framework and provides empirical evidence that DenseNet121-based transfer learning offers an effective and stable solution for household appliance image classification. The findings provide a reliable technical foundation for the future development of vision-based energy-aware applications.

Keywords—deep learning, DenseNet121, energy-aware systems, household appliance recognition, image classification, transfer learning

[PDF]

Cite: Fuzy Yustika Manik, Anandhini Medianty Nababan, and T. M. Rezha Taufiqurrahman, "Comparative Evaluation of Transfer Learning Models for Household Appliance Image Classification," International Journal of Computer Theory and Engineering, vol. 18, no. 3, pp. 208-217, 2026.

Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).