PC Component Classification Using Convolutional Neural Network MobileNetV2 with Transfer Learning
DOI:
https://doi.org/10.55927/marcopolo.v3i7.82Keywords:
PC Components, Image Classification, MobileNetV2, Transfer Learning, Deep LearningAbstract
This research successfully developed and evaluated a computer vision-based classification system for PC components using the MobileNetV2 Convolutional Neural Network (CNN) architecture with transfer learning. A diverse dataset comprising 14 categories of PC components was collected and processed through stratified data splitting (70% training, 15% validation, 15% testing) and extensive data augmentation to enhance model robustness. The MobileNetV2 model, initialized with ImageNet pre-trained weights and fine-tuned with a custom classification layer, demonstrated effective convergence over 30 epochs. The model achieved a peak validation accuracy of 0.9143 and a final test accuracy of 92.38% with a loss of 0.2890. Performance analysis per class, using a classification report, indicated strong results, with most classes exhibiting precision, recall, and F1-scores above 0.90. Specifically, classes like keyboard, speakers, and webcam achieved perfect recall (1.00). While the confusion matrix revealed some misclassifications, particularly for 'case' and 'hdd' components, the overall performance is highly satisfactory, confirming the feasibility of MobileNetV2 for accurate PC component classification. This study contributes to automating PC component identification in industrial and technical contexts, paving the way for more efficient quality control, inventory management, and diagnostics.
References
I. A. Soomro, A. Ahmad, dan R. H. Raza, “Printed Circuit Board identification using Deep Convolutional Neural Networks to facilitate recycling,” Resour Conserv Recycl, vol. 177, Feb 2022, doi: 10.1016/j.resconrec.2021.105963.
K. D. Weerakkody, R. Balasundaram, E. Osagie, dan J. Alshehabi Al-Ani, “Automated Defect Identification System in Printed Circuit Boards Using Region-Based Convolutional Neural Networks,” Electronics (Switzerland), vol. 14, no. 8, Apr 2025, doi: 10.3390/electronics14081542.
M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, dan L.-C. Chen, “MobileNetV2: Inverted Residuals and Linear Bottlenecks,” 2018. doi: 10.48550/arXiv.1704.04861.
M. Andreetto Google Inc dan A. G. Howard Menglong Zhu Bo Chen Dmitry Kalenichenko Weijun Wang Tobias Weyand Marco Andreetto, “MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications,” 2017, doi: 10.48550/arXiv.1704.04861.
F. Uysal, “Electronic Components Detection Using Various Deep Learning Based Neural Network Models,” International Journal of Computational and Experimental Science and Engineering, vol. 11, no. 1, hlm. 542–549, 2024, doi: 10.22399/ijcesen.855.
A. Osmani, T. Rahman, dan S. Islam, Voltavision: A Transfer Learning Model For Electronic Component Classification. 2024.
S. Hożyń, “Convolutional Neural Networks for Classifying Electronic Components in Industrial Applications,” Energies (Basel), vol. 16, no. 2, Jan 2023, doi: 10.3390/en16020887.
C. Surmeli dan H. Ekenel, An Efficient Vision Transformer Model for PCB Component Classification. 2023. doi: 10.1109/SIU59756.2023.10224039.
D. Varna dan V. Abromavičius, “A System for a Real-Time Electronic Component Detection and Classification on a Conveyor Belt,” Applied Sciences (Switzerland), vol. 12, no. 11, Jun 2022, doi: 10.3390/app12115608.
R. Huang, J. Gu, X. Sun, H. Yongtao, dan S. Uddin, “A Rapid Recognition Method for Electronic Components Based on the Improved YOLO-V3 Network,” Electronics (Basel), vol. 8, hlm. 825, Jul 2019, doi: 10.3390/electronics8080825.
M. Mohsin, S. Rovetta, F. Masulli, dan A. Cabri, Real-Time Detection of Electronic Components in Waste Printed Circuit Boards: A Transformer-Based Approach. 2024. doi: 10.48550/arXiv.2409.16496.
A. S. Razavian, H. Azizpour, J. Sullivan, dan S. Carlsson, “CNN Features Off-the-Shelf: An Astounding Baseline for Recognition,” dalam 2014 IEEE Conference on Computer Vision and Pattern Recognition Workshops, 2014, hlm. 512–519. doi: 10.1109/CVPRW.2014.131.
